diff --git a/licence.txt b/LICENCE.txt
similarity index 100%
rename from licence.txt
rename to LICENCE.txt
diff --git a/README.md b/README.md
index cda4897e208c55ec8ff3a2d7a67e5454380690a5..3cc84a35c0f680b71459928e018e4a454ba8b3c5 100644
--- a/README.md
+++ b/README.md
@@ -1,5 +1,83 @@
# dlc_flytracker
+Python package for the tracking the 3d kinematics of animals (currently only flies) build around [DeepLabCut](https://github.com/DeepLabCut/DeepLabCut)
+
+## Introduction
+The package allow the fitting of an animal (chosen) 3d skeleton to the tracking results from DeepLabCut.
+The usual steps:
+- **Post-processing of video recordings (at least 2 cameras):** image enhancement, dynamic cropping around moving animal and the stitching of multiple camera views (to improve tracking with DeepLabCut)
+- **Define the joints of the animal skeleton** that will be used for the tracking. You can use a previously made skeleton (e.g. flies), or make your own from body modules (see the section ??? below to see an example).
+- **(Outside this package) Training of a DeepLabCut (DLC) network:** usually done using the DLC user interface. Require to manually label 100-200 images. See the section ??? below to see an example, and how DLC settings can be tweaked in order to get good tracking results with stitched images.
+- **Run DLC tracking on the full dataset**
+- **Load the DLC results in dlc_flytracker**
+- **Reconstruct 3d coordinates of the skeleton joints** from their 2d coordinates tracked with DeepLabCut
+- **Optimize fit the 3d skeleton** to the 2d/3d coordinates of the joints
+- **Post-processing of the kinematics parameters**
+
+
+## Citation
+???
+
+
+## Getting Started
+### Prerequisite
+dlc_flytracker requires `deeplabcut`, `tensorflow` and other common packages like `numpy`, `scipy`, `matplotlib`. For a full list, you can check the file [requirements.txt](requirements.txt)
+
+### Installing
+???
+
+
+## Usage
+Here the main steps introduced earlier are explained in details with examples.
+
+### Post-processing of video recordings
+??? (at least 2 cameras):
+- **Image enhancement:** ???
+- **Dynamic cropping around moving animal:** ???
+- **Stitching of multiple camera views:** ??? (to improve tracking with DeepLabCut)
+
+
+
+
+### Build animal skeleton
+???
+
+#### in DeepLabCut
+???
+
+#### in dlc_flytracker
+???
+
+### Training a network in DeepLabCut
+- ??? using the DLC user interface. Require to manually label 100-200 images. See the section ??? below to see an example, and how DLC settings can be tweaked in order to get good tracking results with stitched images.
+
+### Run DLC tracking on the full dataset
+???
+
+### Load the DLC results in dlc_flytracker
+???
+
+### Reconstruct 3d coordinates of the skeleton joints from their 2d coordinates tracked with DeepLabCut
+???
+
+### Optimize fit the 3d skeleton to the 2d/3d coordinates of the joints
+???
+
+### Post-processing of the kinematics parameters
+???
+
+## Notes
+### Running the tests
+To run tests locally, type:
+```bash
+> pytest dlc_flytracker
+```
+
+
+## License
+
+This project utilizes the [LGPL LICENSE](LICENCE.txt).
+Allows developers and companies to use and integrate a software component released under the LGPL into their own (even proprietary) software without being required by the terms of a strong copyleft license to release the source code of their own components. However, any developer who modifies an LGPL-covered component is required to make their modified version available under the same LGPL license.
\ No newline at end of file
diff --git a/cam/__init__.py b/cam/__init__.py
deleted file mode 100644
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..0000000000000000000000000000000000000000
diff --git a/cam/calib.py b/cam/calib.py
deleted file mode 100644
index 118bdddb8b3db29294dec0573bc85b5c7e2ded47..0000000000000000000000000000000000000000
--- a/cam/calib.py
+++ /dev/null
@@ -1,372 +0,0 @@
-"""
-I found this code at https://www.mail-archive.com/floatcanvas@mithis.com/msg00513.html
-(author: Marcos Duarte) and I made several small modifications (me: Antoine Cribellier)
-
-
-Camera calibration and point reconstruction based on direct linear transformation (DLT).
-
-The fundamental problem here is to find a mathematical relationship between the
- coordinates of a 3D point and its projection onto the images plane. The DLT
- (a linear apporximation to this problem) is derived from modelling the object
- and its projection on the images plane as a pinhole camera situation.
-In simplistic terms, using the pinhole camera model, it can be found by similar
- triangles the following relation between the images coordinates (u,v) and the 3D
- point (X,Y,Z):
- [ u ] [ L1 L2 L3 L4 ] [ X ]
- [ v ] = [ L5 L6 L7 L8 ] [ Y ]
- [ 1 ] [ L9 L10 L11 L12 ] [ Z ]
- [ 1 ]
-The matrix L is kwnown as the camera matrix or camera projection matrix. For a
- 2D point (X,Y), the last column of the matrix doesn't exist. In fact, the L12
- term (or L9 for 2D DLT) is not independent from the other parameters and then
- there are only 11 (or 8 for 2D DLT) independent parameters in the DLT to be
- determined.
-
-DLT is typically used in two steps: 1. camera calibration and 2. object (point)
- reconstruction.
-The camera calibration step consists in digitizing points with known coordiantes
- in the real space.
-At least 4 points are necessary for the calibration of a plane (2D DLT) and at
- least 6 points for the calibration of a volume (3D DLT). For the 2D DLT, at least
- one view of the object (points) must be entered. For the 3D DLT, at least 2
- different views of the object (points) must be entered.
-These coordinates (from the object and images(s)) are inputed to the calib
- algorithm which estimates the camera parameters (8 for 2D DLT and 11 for 3D DLT).
-With these camera parameters and with the camera(s) at the same position of the
- calibration step, we now can reconstruct the real position of any point inside
- the calibrated space (area for 2D DLT and volume for the 3D DLT) from the point
- position(s) viewed by the same fixed camera(s).
-
-This code can perform 2D or 3D DLT with any number of views (cameras).
-For 3D DLT, at least two views (cameras) are necessary.
-
-There are more accurate (but more complex) algorithms for camera calibration that
- also consider lens distortion. For example, OpenCV and Tsai softwares have been
- ported to Python. However, DLT is classic, simple, and effective (fast) for
- most applications.
-
-About DLT, see: http://kwon3d.com/theory/dlt/html
-
-This code is based on different implementations and teaching material on DLT
- found in the internet.
-"""
-
-import numpy as np
-
-
-def calib(nd, xyz, uv):
- """
- Camera calibration by DLT using known object points and their images points.
-
- This code performs 2D or 3D DLT camera calibration with any number of views (cameras).
- For 3D DLT, at least two views (cameras) are necessary.
-
- The coordinates (x,y,z and u,v) are given as columns and the different points as rows.
- For the 2D DLT (object planar space), only the first 2 columns (x and y) are used.
- There must be at least 6 calibration points for the 3D DLT and 4 for the 2D
-
- Args:
- nd: the number of dimensions of the object space: 3 for 3D DLT and 2 for 2D
- xyz: the coordinates in the object 3D or 2D space of the calibration points (3*N).
- uv: the coordinates in the images 2D space of these calibration points (2*N).
-
- Returns:
- L: array of the 8 or 11 parameters of the calibration matrix (9*1 or 12*1)
- err: error of the DLT (mean residual of the DLT transformation in units of camera coordinates).
- """
-
- # Convert all variables to numpy array:
- xyz = np.asarray(xyz)
- uv = np.asarray(uv)
- # number of points:
- nb_pts = xyz.shape[0]
- # Check the parameters:
- if uv.shape[0] != nb_pts:
- raise ValueError('xyz (%d points) and uv (%d points) have different number of points.' % (nb_pts, uv.shape[0]))
- if (nd == 2 and xyz.shape[1] != 2) or (nd == 3 and xyz.shape[1] != 3):
- raise ValueError('Incorrect number of coordinates (%d) for %dD DLT (it should be %d).' % (xyz.shape[1], nd, nd))
- if nd == 3 and nb_pts < 6 or nd == 2 and nb_pts < 4:
- raise ValueError(
- '%dD DLT requires at least %d calibration points. Only %d points were entered.' % (nd, 2 * nd, nb_pts))
-
- # Normalize the data to improve the DLT quality (DLT is dependent of the system of coordinates).
- # This is relevant when there is a considerable perspective distortion.
- # Normalization: mean position at origin and mean distance equals to 1 at each direction.
- Txyz, xyzn = Normalization(nd, xyz)
- Tuv, uvn = Normalization(2, uv)
-
- A = []
- if nd == 2: # 2D DLT
- for i in range(nb_pts):
- x, y = xyzn[i, 0], xyzn[i, 1]
- u, v = uvn[i, 0], uvn[i, 1]
- A.append([x, y, 1, 0, 0, 0, -u * x, -u * y, -u])
- A.append([0, 0, 0, x, y, 1, -v * x, -v * y, -v])
- elif nd == 3: # 3D DLT
- for i in range(nb_pts):
- x, y, z = xyzn[i, 0], xyzn[i, 1], xyzn[i, 2]
- u, v = uvn[i, 0], uvn[i, 1]
- A.append([x, y, z, 1, 0, 0, 0, 0, -u * x, -u * y, -u * z, -u])
- A.append([0, 0, 0, 0, x, y, z, 1, -v * x, -v * y, -v * z, -v])
-
- # convert A to array
- A = np.asarray(A)
- # Find the 11 (or 8 for 2D DLT) parameters:
- U, S, Vh = np.linalg.svd(A)
- # The parameters are in the last line of Vh and normalize them:
- L = Vh[-1, :] / Vh[-1, -1]
- # Camera projection matrix:
- H = L.reshape(3, nd + 1)
- # Denormalization:
- H = np.dot(np.dot(np.linalg.pinv(Tuv), H), Txyz);
- H = H / H[-1, -1]
- L = H.flatten('C') # was L = H.flatten(0)
- # Mean error of the DLT (mean residual of the DLT transformation in units of camera coordinates):
- uv2 = np.dot(H, np.concatenate((xyz.T, np.ones((1, xyz.shape[0])))))
- uv2 = uv2 / uv2[2, :]
- # mean distance:
- err = np.sqrt(np.mean(np.sum((uv2[0:2, :].T - uv) ** 2, 1)))
-
- return L, err
-
-
-def find2d(nc, Ls, xyz):
- """
- Find 2D reprojection of object point to images point based on the DLT parameters.
-
- Args:
- nc: number of cameras (views) used.
- Ls: (array type) are the camera calibration parameters of each camera (is the output of calib function).
- The Ls parameters are given as columns and the Ls for different cameras as rows (nc*9 or nc*12).
- xyz: coordinates in the object 3D or 2D space of the calibration points (3*N).
-
- Returns:
- uv: coordinates in the images 2D space of these calibration points.
- The coordinates (x,y,z and u,v) are given as columns and the different points as rows.
- For the 2D DLT (object planar space), only the first 2 columns (x and y) are used.
- There must be at least 6 calibration points for the 3D DLT and 4 for the 2D
- """
-
- # Convert Ls to array:
- Ls = np.asarray(Ls)
-
- if nc == 1:
- H = Ls.reshape((3, 4))
-
- uv = np.dot(H, np.concatenate((xyz.T, np.ones((1, xyz.shape[0])))))
- uv = uv / uv[2, :]
- uvs = uv[0:2, :].T
-
- else:
- uvs = np.array([])
- for i in range(nc):
- L = Ls[i, :]
- H = L.reshape((3, 4))
-
- uv = np.dot(H, np.concatenate((xyz.T, np.ones((1, xyz.shape[0])))))
- uv = uv / uv[2, :]
- uv = uv[0:2, :].T
-
- uvs = np.vstack([uvs, uv]) if uvs.size else uv
-
- return uvs
-
-
-def recon3D(nd, nc, Ls, uvs):
- """
- Reconstruction of object point from images point(s) based on the DLT parameters.
-
- This code performs 2D or 3D DLT point reconstruction with any number of views (cameras).
- For 3D DLT, at least two views (cameras) are necessary.
-
- Args:
- nd: number of dimensions of the object space: 3 for 3D DLT and 2 for 2D
- nc: number of cameras (views) used.
- Ls: (array type) are the camera calibration parameters of each camera (is the output of calib function).
- The Ls parameters are given as columns and the Ls for different cameras as rows (nc*9 or nc*12).
- uvs: coordinates of the point in the images 2D space of each camera
- The coordinates of the point are given as columns and the different views as rows (2*N).
-
- Returns:
- xyz: point coordinates in space (3*N)
- """
-
- # Convert Ls to array:
- Ls = np.asarray(Ls)
-
- # Check the parameters:
- if np.isnan(uvs).any():
- raise ValueError("uvs shouldn't contain NaN!")
- if Ls.ndim == 1 and nc != 1:
- raise ValueError(
- 'Number of views (%d) and number of sets of camera calibration parameters (1) are different.' % (nc))
- if Ls.ndim > 1 and nc != Ls.shape[0]:
- raise ValueError(
- 'Number of views (%d) and number of sets of camera calibration parameters (%d) are different.' % (
- nc, Ls.shape[0]))
- if nd == 3 and Ls.ndim == 1:
- raise ValueError('At least two sets of camera calibration parameters are needed for 3D point reconstruction.')
-
- if nc == 1: # 2D and 1 camera (view), the simplest (and fastest) case
- # One could calculate inv(H) and input that to the code to speed up things if needed.
- # (If there is only 1 camera, this transformation is all Floatcanvas2 might need)
- Hinv = np.linalg.inv(Ls.reshape(3, 3))
- # Point coordinates in space:
- xyz = np.dot(Hinv, [uvs[0], uvs[1], 1])
- xyz = xyz[0:2] / xyz[2]
-
- else:
- M = []
- for i in range(nc):
- L = Ls[i, :]
- u, v = uvs[i][0], uvs[i][1] # this indexing works for both list and numpy array
- if nd == 2:
- M.append([L[0] - u * L[6], L[1] - u * L[7], L[2] - u * L[8]])
- M.append([L[3] - v * L[6], L[4] - v * L[7], L[5] - v * L[8]])
- elif nd == 3:
- M.append([L[0] - u * L[8], L[1] - u * L[9], L[2] - u * L[10], L[3] - u * L[11]])
- M.append([L[4] - v * L[8], L[5] - v * L[9], L[6] - v * L[10], L[7] - v * L[11]])
-
-
- # Find the xyz coordinates:
- U, S, Vh = np.linalg.svd(np.asarray(M))
- # Point coordinates in space:
- xyz = Vh[-1, 0:-1] / Vh[-1, -1]
-
- return xyz
-
-
-def Normalization(nd, x):
- """
- Normalization of coordinates (centroid to the origin and mean distance of sqrt(2 or 3).
-
- Args:
- nd: number of dimensions (2 for 2D; 3 for 3D)
- x: data to be normalized (directions at different columns and points at rows)
-
- Returns:
- Tr: the transformation matrix (translation plus scaling)
- x: the transformed data
- """
-
- x = np.asarray(x)
- m, s = np.mean(x, 0), np.std(x)
- if nd == 2:
- Tr = np.array([[s, 0, m[0]], [0, s, m[1]], [0, 0, 1]])
- else:
- Tr = np.array([[s, 0, 0, m[0]], [0, s, 0, m[1]], [0, 0, s, m[2]], [0, 0, 0, 1]])
-
- Tr = np.linalg.inv(Tr)
- x = np.dot(Tr, np.concatenate((x.T, np.ones((1, x.shape[0])))))
- x = x[0:nd, :].T
-
- return Tr, x
-
-
-if __name__ == '__main__':
- # Tests of DLTx
- print('')
- print('Test of camera calibration and point reconstruction based on direct linear transformation (DLT).')
- print('3D (x, y, z) coordinates (in cm) of the corner of a cube (the measurement error is at least 0.2 cm):')
- xyz = [[0, 0, 0], [0, 12.3, 0], [14.5, 12.3, 0], [14.5, 0, 0], [0, 0, 14.5], [0, 12.3, 14.5], [14.5, 12.3, 14.5],
- [14.5, 0, 14.5]]
- print(np.asarray(xyz))
- print('2D (u, v) coordinates (in pixels) of 4 different views of the cube:')
- uv1 = [[1302, 1147], [1110, 976], [1411, 863], [1618, 1012], [1324, 812], [1127, 658], [1433, 564], [1645, 704]]
- uv2 = [[1094, 1187], [1130, 956], [1514, 968], [1532, 1187], [1076, 854], [1109, 647], [1514, 659], [1523, 860]]
- uv3 = [[1073, 866], [1319, 761], [1580, 896], [1352, 1016], [1064, 545], [1304, 449], [1568, 557], [1313, 668]]
- uv4 = [[1205, 1511], [1193, 1142], [1601, 1121], [1631, 1487], [1157, 1550], [1139, 1124], [1628, 1100],
- [1661, 1520]]
- print('uv1:')
- print(np.asarray(uv1))
- print('uv2:')
- print(np.asarray(uv2))
- print('uv3:')
- print(np.asarray(uv3))
- print('uv4:')
- print(np.asarray(uv4))
-
- print('')
- print('Use 4 views to perform a 3D calibration of the camera with 8 points of the cube:')
- nd = 3
- nc = 4
- L1, err1 = calib(nd, xyz, uv1)
- print('Camera calibration parameters based on view #1:')
- print(L1)
- print('Error of the calibration of view #1 (in pixels):')
- print(err1)
- L2, err2 = calib(nd, xyz, uv2)
- print('Camera calibration parameters based on view #2:')
- print(L2)
- print('Error of the calibration of view #2 (in pixels):')
- print(err2)
- L3, err3 = calib(nd, xyz, uv3)
- print('Camera calibration parameters based on view #3:')
- print(L3)
- print('Error of the calibration of view #3 (in pixels):')
- print(err3)
- L4, err4 = calib(nd, xyz, uv4)
- print('Camera calibration parameters based on view #4:')
- print(L4)
- print('Error of the calibration of view #4 (in pixels):')
- print(err4)
- xyz1234 = np.zeros((len(xyz), 3))
- L1234 = [L1, L2, L3, L4]
- for i in range(len(uv1)):
- xyz1234[i, :] = recon3D(nd, nc, L1234, [uv1[i], uv2[i], uv3[i], uv4[i]])
- print('Reconstruction of the same 8 points based on 4 views and the camera calibration parameters:')
- print(xyz1234)
- print('Mean error of the point reconstruction using the DLT (error in cm):')
- print(np.mean(np.sqrt(np.sum((np.array(xyz1234) - np.array(xyz)) ** 2, 1))))
-
- print('')
- print('Test of the 2D DLT')
- print('2D (x, y) coordinates (in cm) of the corner of a square (the measurement error is at least 0.2 cm):')
- xy = [[0, 0], [0, 12.3], [14.5, 12.3], [14.5, 0]]
- print(np.asarray(xy))
- print('2D (u, v) coordinates (in pixels) of 2 different views of the square:')
- uv1 = [[1302, 1147], [1110, 976], [1411, 863], [1618, 1012]]
- uv2 = [[1094, 1187], [1130, 956], [1514, 968], [1532, 1187]]
- print('uv1:')
- print(np.asarray(uv1))
- print('uv2:')
- print(np.asarray(uv2))
- print('')
- print('Use 2 views to perform a 2D calibration of the camera with 4 points of the square:')
- nd = 2
- nc = 2
- L1, err1 = calib(nd, xy, uv1)
- print('Camera calibration parameters based on view #1:')
- print(L1)
- print('Error of the calibration of view #1 (in pixels):')
- print(err1)
- L2, err2 = calib(nd, xy, uv2)
- print('Camera calibration parameters based on view #2:')
- print(L2)
- print('Error of the calibration of view #2 (in pixels):')
- print(err2)
- xy12 = np.zeros((len(xy), 2))
- L12 = [L1, L2]
- for i in range(len(uv1)):
- xy12[i, :] = recon3D(nd, nc, L12, [uv1[i], uv2[i]])
- print('Reconstruction of the same 4 points based on 2 views and the camera calibration parameters:')
- print(xy12)
- print('Mean error of the point reconstruction using the DLT (error in cm):')
- print(np.mean(np.sqrt(np.sum((np.array(xy12) - np.array(xy)) ** 2, 1))))
-
- print('')
- print('Use only one view to perform a 2D calibration of the camera with 4 points of the square:')
- nd = 2
- nc = 1
- L1, err1 = calib(nd, xy, uv1)
- print('Camera calibration parameters based on view #1:')
- print(L1)
- print('Error of the calibration of view #1 (in pixels):')
- print(err1)
- xy1 = np.zeros((len(xy), 2))
- for i in range(len(uv1)):
- xy1[i, :] = recon3D(nd, nc, L1, uv1[i])
- print('Reconstruction of the same 4 points based on one view and the camera calibration parameters:')
- print(xy1)
- print('Mean error of the point reconstruction using the DLT (error in cm):')
- print(np.mean(np.sqrt(np.sum((np.array(xy1) - np.array(xy)) ** 2, 1))))
diff --git a/camera/__init__.py b/camera/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e5be388838716757e15b8c07a8e8cf4629092b20
--- /dev/null
+++ b/camera/__init__.py
@@ -0,0 +1 @@
+import camera.calib as calib
diff --git a/camera/calib.py b/camera/calib.py
new file mode 100644
index 0000000000000000000000000000000000000000..09c5ab70f12d2d396b43ddf780e0c2c980c2baee
--- /dev/null
+++ b/camera/calib.py
@@ -0,0 +1,460 @@
+"""
+I (Antoine Cribellier) found this code at https://www.mail-archive.com/floatcanvas@mithis.com/msg00513.html
+(author: Marcos Duarte)
+I made various modification to clarify things and added functions (read_dlt_coefs, save_dlt_coefs, gen_dlt_coefs)
+Also the functions are now working use 8 or 11 DLT coefficients instead of 9 or 12 before (the last one was always 1)
+
+
+Camera calibration and point reconstruction based on direct linear transformation (DLT).
+
+The fundamental problem here is to find a mathematical relationship between the
+ coordinates of a 3D point and its projection onto the images plane. The DLT
+ (a linear apporximation to this problem) is derived from modelling the object
+ and its projection on the images plane as a pinhole camera situation.
+In simplistic terms, using the pinhole camera model, it can be found by similar
+ triangles the following relation between the images coordinates (u,v) and the 3D
+ point (X,Y,Z):
+ [ u ] [ L1 L2 L3 L4 ] [ X ]
+ [ v ] = [ L5 L6 L7 L8 ] [ Y ]
+ [ 1 ] [ L9 L10 L11 L12 ] [ Z ]
+ [ 1 ]
+The matrix dlt_coef is kwnown as the camera matrix or camera projection matrix. For a
+ 2D point (X,Y), the last column of the matrix doesn't exist. In fact, the L12
+ term (or L9 for 2D DLT) is not independant from the other parameters and then
+ there are only 11 (or 8 for 2D DLT) independent parameters in the DLT to be
+ determined.
+
+DLT is typically used in two steps: 1. camera calibration and 2. object (point)
+ reconstruction.
+The camera calibration step consists in digitizing points with known coordiantes
+ in the real space.
+At least 4 points are necessary for the calibration of a plane (2D DLT) and at
+ least 6 points for the calibration of a volume (3D DLT). For the 2D DLT, at least
+ one view of the object (points) must be entered. For the 3D DLT, at least 2
+ different views of the object (points) must be entered.
+These coordinates (from the object and images(s)) are inputed to the calib
+ algorithm which estimates the camera parameters (8 for 2D DLT and 11 for 3D DLT).
+With these camera parameters and with the camera(s) at the same position of the
+ calibration step, we now can reconstruct the real position of any point inside
+ the calibrated space (area for 2D DLT and volume for the 3D DLT) from the point
+ position(s) viewed by the same fixed camera(s).
+
+This code can perform 2D or 3D DLT with any number of views (cameras).
+For 3D DLT, at least two views (cameras) are necessary.
+
+There are more accurate (but more complex) algorithms for camera calibration that
+ also consider lens distortion. For example, OpenCV and Tsai softwares have been
+ ported to Python. However, DLT is classic, simple, and effective (fast) for
+ most applications.
+
+About DLT, see: http://kwon3d.com/theory/dlt/html
+
+This code is based on different implementations and teaching material on DLT
+ found in the internet.
+"""
+from typing import List
+
+import numpy as np
+from matplotlib import pyplot as plt
+
+
+def calib(nb_dim, xyz, uv):
+ """
+ Camera calibration by DLT using known object points and their images points.
+
+ This code performs 2D or 3D DLT camera calibration with any number of views (cameras).
+ For 3D DLT, at least two views (cameras) are necessary.
+
+ The coordinates (x,y,z and u,v) are given as columns and the different points as rows.
+ For the 2D DLT (object planar space), only the first 2 columns (x and y) are used.
+ There must be at least 6 calibration points for the 3D DLT and 4 for the 2D
+
+ Args:
+ nb_dim: the number of dimensions of the object space: 3 for 3D DLT and 2 for 2D
+ xyz: the coordinates in the object 3D or 2D space of the calibration points (3*N).
+ uv: the coordinates in the images 2D space of these calibration points (2*N).
+
+ Returns:
+ dlt_coef: array of the 8 or 11 parameters of the calibration matrix (8*1 or 11*1)
+ repro_err: error of the DLT (mean residual of the DLT transformation in units of camera coordinates).
+ """
+
+ # Convert all variables to numpy array:
+ xyz = np.asarray(xyz)
+ uv = np.asarray(uv)
+
+ # Number of points:
+ nb_pts = xyz.shape[0]
+
+ # Check the parameters:
+ if uv.shape[0] != nb_pts:
+ raise ValueError('xyz (%d points) and uv (%d points) have different number of points.' % (nb_pts, uv.shape[0]))
+
+ if (nb_dim == 2 and xyz.shape[1] != 2) or (nb_dim == 3 and xyz.shape[1] != 3):
+ raise ValueError('Incorrect number of coordinates (%d) for %dD DLT (it should be %d).' % (xyz.shape[1], nb_dim, nb_dim))
+
+ if nb_dim == 3 and nb_pts < 6 or nb_dim == 2 and nb_pts < 4:
+ raise ValueError(
+ '%dD DLT requires at least %d calibration points. Only %d points were entered.' % (nb_dim, 2 * nb_dim, nb_pts))
+
+ # Normalize the data to improve the DLT quality (DLT is dependent of the system of coordinates).
+ # This is relevant when there is a considerable perspective distortion.
+ # Normalization: mean position at origin and mean distance equals to 1 at each direction.
+ Txyz, xyzn = normalization(nb_dim, xyz)
+ Tuv, uvn = normalization(2, uv)
+
+ A = []
+ if nb_dim == 2: # 2D DLT
+ for i in range(nb_pts):
+ x, y = xyzn[i, 0], xyzn[i, 1]
+ u, v = uvn[i, 0], uvn[i, 1]
+ A.append([x, y, 1, 0, 0, 0, -u * x, -u * y, -u])
+ A.append([0, 0, 0, x, y, 1, -v * x, -v * y, -v])
+ elif nb_dim == 3: # 3D DLT
+ for i in range(nb_pts):
+ x, y, z = xyzn[i, 0], xyzn[i, 1], xyzn[i, 2]
+ u, v = uvn[i, 0], uvn[i, 1]
+ A.append([x, y, z, 1, 0, 0, 0, 0, -u * x, -u * y, -u * z, -u])
+ A.append([0, 0, 0, 0, x, y, z, 1, -v * x, -v * y, -v * z, -v])
+
+ # Convert A to array
+ A = np.asarray(A)
+
+ # Find the 11 (or 8 for 2D DLT) parameters:
+ U, S, Vh = np.linalg.svd(A)
+
+ # The parameters are in the last line of Vh and normalize them:
+ dlt_coef = Vh[-1, :] / Vh[-1, -1]
+
+ # Camera projection matrix:
+ H = dlt_coef.reshape(3, nb_dim + 1)
+
+ # De-normalization:
+ H = np.dot(np.dot(np.linalg.pinv(Tuv), H), Txyz);
+ H = H / H[-1, -1]
+ dlt_coef = H.flatten('C') # was dlt_coef = H.flatten(0)
+
+ dlt_coef = dlt_coef[:-1] # to remove last coefficient as it's not independent from the other (should always be 1.0)
+
+ # Mean error of the DLT (mean residual of the DLT transformation in units of camera coordinates):
+ uv2 = np.dot(H, np.concatenate((xyz.T, np.ones((1, xyz.shape[0])))))
+ uv2 = uv2 / uv2[2, :]
+
+ # Mean distance:
+ repro_err = np.sqrt(np.mean(np.sum((uv2[0:2, :].T - uv) ** 2, 1)))
+
+ return dlt_coef, repro_err
+
+
+def find2d(nb_cam, dlt_coefs, xyz):
+ """
+ Find 2D reprojection of object point to images point based on the DLT parameters.
+
+ Args:
+ nb_cam: number of cameras (views) used.
+ dlt_coefs: (array type) are the camera calibration parameters of each camera (is the output of calib function).
+ The DLT coefficients parameters are given as columns with the cameras as rows (nb_cam*8 or nb_cam*11).
+ xyz: coordinates in the object 3D or 2D space of the calibration points (3*N).
+
+ Returns:
+ uv: coordinates in the images 2D space of these calibration points.
+ The coordinates (x,y,z and u,v) are given as columns and the different points as rows.
+ For the 2D DLT (object planar space), only the first 2 columns (x and y) are used.
+ There must be at least 6 calibration points for the 3D DLT and 4 for the 2D
+ """
+
+ dlt_coefs = np.asarray(dlt_coefs) # Convert dlt_coefs to array:
+
+ if nb_cam == 1:
+ dlt_coefs = np.append(dlt_coefs, 1.0) # to allow reshaping to 3*4 matrix
+ H = dlt_coefs.reshape((3, 4))
+
+ uv = np.dot(H, np.concatenate((xyz.T, np.ones((1, xyz.shape[0])))))
+ uv = uv / uv[2, :]
+ uvs = uv[0:2, :].T
+
+ else:
+ uvs = np.array([])
+ for i in range(nb_cam):
+ dlt_coef = dlt_coefs[i, :]
+ dlt_coef = np.append(dlt_coef, 1.0) # to allow reshaping to 3*4 matrix
+
+ H = dlt_coef.reshape((3, 4))
+
+ uv = np.dot(H, np.concatenate((xyz.T, np.ones((1, xyz.shape[0])))))
+ uv = uv / uv[2, :]
+ uv = uv[0:2, :].T
+
+ uvs = np.vstack([uvs, uv]) if uvs.size else uv
+
+ return uvs
+
+
+def recon3d(nb_dim, nb_cam, dlt_coefs, uvs):
+ """
+ Reconstruction of object point from images point(s) based on the DLT parameters.
+
+ This code performs 2D or 3D DLT point reconstruction with any number of views (cameras).
+ For 3D DLT, at least two views (cameras) are necessary.
+
+ Args:
+ nb_dim: number of dimensions of the object space: 3 for 3D DLT and 2 for 2D
+ nb_cam: number of cameras (views) used.
+ dlt_coefs: (array type) are the camera calibration parameters of each camera (is the output of calib function).
+ The dlt_coefs parameters are given as columns with the cameras as rows (nb_cam*8 or nb_cam*11).
+ uvs: coordinates of the point in the images 2D space of each camera
+ The coordinates of the point are given as columns and the different views as rows (2*N).
+
+ Returns:
+ xyz: point coordinates in space (3*N)
+ """
+
+ dlt_coefs = np.asarray(dlt_coefs) # Convert dlt_coefs to array:
+
+ # Check the parameters:
+ if np.isnan(uvs).any():
+ raise ValueError("uvs shouldn't contain NaN!")
+
+ if dlt_coefs.ndim == 1 and nb_cam != 1:
+ raise ValueError(
+ 'Number of views (%d) and number of sets of camera calibration parameters (1) are different.' % (nb_cam))
+
+ if dlt_coefs.ndim > 1 and nb_cam != dlt_coefs.shape[0]:
+ raise ValueError(
+ 'Number of views (%d) and number of sets of camera calibration parameters (%d) are different.' % (
+ nb_cam, dlt_coefs.shape[0]))
+
+ if nb_dim == 3 and dlt_coefs.ndim == 1:
+ raise ValueError('At least two sets of camera calibration parameters are needed for 3D point reconstruction.')
+
+ if nb_cam == 1: # 2D and 1 camera (view), the simplest (and fastest) case
+ dlt_coefs = np.append(dlt_coefs, 1.0) # to allow reshaping to 3*3 matrices
+
+ # One could calculate inv(H) and input that to the code to speed up things if needed.
+ # (If there is only 1 camera, this transformation is all Floatcanvas2 might need)
+ Hinv = np.linalg.inv(dlt_coefs.reshape(3, 3))
+
+ # Point coordinates in space:
+ xyz = np.dot(Hinv, [uvs[0], uvs[1], 1])
+ xyz = xyz[0:2] / xyz[2]
+
+ else:
+ M = []
+ for i in range(nb_cam):
+ dlt_coef = dlt_coefs[i, :]
+ dlt_coef = np.append(dlt_coef, 1.0)
+
+ u, v = uvs[i][0], uvs[i][1] # this indexing works for both list and numpy array
+ if nb_dim == 2:
+ M.append([dlt_coef[0] - u * dlt_coef[6], dlt_coef[1] - u * dlt_coef[7], dlt_coef[2] - u * dlt_coef[8]])
+ M.append([dlt_coef[3] - v * dlt_coef[6], dlt_coef[4] - v * dlt_coef[7], dlt_coef[5] - v * dlt_coef[8]])
+
+ elif nb_dim == 3:
+ M.append([dlt_coef[0] - u * dlt_coef[8], dlt_coef[1] - u * dlt_coef[9], dlt_coef[2] - u * dlt_coef[10], dlt_coef[3] - u * dlt_coef[11]])
+ M.append([dlt_coef[4] - v * dlt_coef[8], dlt_coef[5] - v * dlt_coef[9], dlt_coef[6] - v * dlt_coef[10], dlt_coef[7] - v * dlt_coef[11]])
+
+ # Find the xyz coordinates:
+ U, S, Vh = np.linalg.svd(np.asarray(M))
+
+ # Point coordinates in space:
+ xyz = Vh[-1, 0:-1] / Vh[-1, -1]
+
+ return xyz
+
+
+def normalization(nb_dim, x):
+ """
+ Normalization of coordinates (centroid to the origin and mean distance of sqrt(2 or 3).
+
+ Args:
+ nb_dim: number of dimensions (2 for 2D; 3 for 3D)
+ x: data to be normalized (directions at different columns and points at rows)
+
+ Returns:
+ Tr: the transformation matrix (translation plus scaling)
+ x: the transformed data
+ """
+
+ x = np.asarray(x)
+ m, s = np.mean(x, 0), np.std(x)
+ if nb_dim == 2:
+ Tr = np.array([[s, 0, m[0]], [0, s, m[1]], [0, 0, 1]])
+ else:
+ Tr = np.array([[s, 0, 0, m[0]], [0, s, 0, m[1]], [0, 0, s, m[2]], [0, 0, 0, 1]])
+
+ Tr = np.linalg.inv(Tr)
+ x = np.dot(Tr, np.concatenate((x.T, np.ones((1, x.shape[0])))))
+ x = x[0:nb_dim, :].T
+
+ return Tr, x
+
+
+def read_dlt_coefs(csv_path: str) -> List[List[float]]:
+ """
+ Read DLT coefficients from .csv file
+
+ Args:
+ csv_path: path of the .csv file where the DLT (11) coefficients (along rows) for all camera (along column) will be written
+
+ Returns:
+ dlt_coefs: List with DLT coefficients (11 * nb_cam)
+ """
+
+ dlt_csv = np.genfromtxt(csv_path, delimiter=',', skip_header=0, names=True)
+ assert len(dlt_csv) == 11
+
+ nb_cam = len(dlt_csv[0])
+ dlt_csv.dtype.names = ['cam{0}'.format(camn) for camn in range(1, nb_cam + 1)]
+
+ dlt_coefs = np.zeros((nb_cam, 11))
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ dlt_coefs[i] = np.array(dlt_csv['cam{0}'.format(camn)])
+
+ return dlt_coefs
+
+
+def save_dlt_coefs(dlt_coefs: List[List[float]], csv_path: str) -> None:
+ """
+ Save DLT coefficients in .csv file
+
+ Args:
+ dlt_coefs: List with DLT coefficients (11 * nb_cam)
+ csv_path: path of the .csv file with DLT (11) coefficients (along rows) for all camera (along column)
+ """
+
+ assert len(dlt_coefs[0]) == 11
+
+ nb_cam = len(dlt_coefs)
+ header = ['cam{0}'.format(camn) for camn in range(1, nb_cam + 1)]
+ header = ','.join(header)
+
+ np.savetxt(csv_path, np.c_[dlt_coefs].T, delimiter=',', header=header)
+
+
+def gen_dlt_coef(obj_coord: List[List[float]], img_pts: List[List[float]]) -> List[List[float]] and List[List[float]]:
+ """
+ Generate DLT coefficients for one camera
+
+ Args:
+ obj_coord: known 3d coordinates of the objects (size: nb_obj*3)
+ img_pts: 2d coordinates for all objects for one camera (size: nb_obj*2)
+
+ Returns:
+ dlt_coef: List of DLT coefficient (11*1) for one camera
+ mean_repro_err: mean reprojcetion error in pixel (or Root-mean-square deviation)
+ """
+
+ dlt_coef, repro_err = calib(3, obj_coord, img_pts)
+
+ return dlt_coef, repro_err
+
+
+def gen_dlt_coefs(obj_coord: List[List[float]], img_pts: List[List[float]],
+ plot_error_analysis: bool = False) -> List[str]:
+ """
+ Generate DLT coefficients for multiple cameras
+
+ Args:
+ obj_coord: known 3d coordinates of the objects (size: nb_obj*3)
+ img_pts: 2d coordinates for all objects for multiple cameras (size: nb_cam*nb_obj*3)
+ plot_error_analysis: Whether or not will do an error analysis
+ (check if not using each 2d point will impact overall calibration error)
+
+ Returns:
+ dlt_coef: List of DLT coefficient (11*nb_cam) for all cameras
+ mean_repro_err: mean reprojection errors in pixel (or Root-mean-square deviation) for all cameras
+ """
+
+ if plot_error_analysis: repro_errs_analysis = {}
+
+ nb_cam = len(obj_coord)
+
+ dlt_coefs = np.zeros((nb_cam, 11))
+ repro_errs = np.zeros((nb_cam))
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+
+ dlt_coefs[i], repro_errs[i] = gen_dlt_coef(obj_coord[i], img_pts[i])
+ print('>> The reprojection error of cam{0} is {1} pixels'.format(camn, repro_errs[i]))
+
+ if plot_error_analysis:
+ repro_errs_analysis[camn] = np.zeros(len(img_pts[i].T[0]))
+
+ for ind_pt in range(len(img_pts[i].T[0])):
+ img_pts_calib_wo_pt = np.array([[s for j, s in enumerate(img_pts[i].T[0]) if j is not ind_pt],
+ [s for j, s in enumerate(img_pts[i].T[0]) if j is not ind_pt]]).T
+ obj_coord_calib_wo_pt = np.array([[s for j, s in enumerate(obj_coord[i].T[0]) if j is not ind_pt],
+ [s for j, s in enumerate(obj_coord[i].T[1]) if j is not ind_pt],
+ [s for j, s in enumerate(obj_coord[i].T[2]) if j is not ind_pt]]).T
+
+ _, repro_errs_analysis[camn][ind_pt] = gen_dlt_coef(obj_coord_calib_wo_pt, img_pts_calib_wo_pt)
+
+ if plot_error_analysis:
+ # Plot all rmse without one point
+
+ for camn in range(1, nb_cam + 1):
+ fig = plt.figure('Error analysis (cam{0})'.format(camn))
+ ax = fig.add_subplot(111)
+ ax.scatter(list(range(1, len(repro_errs_analysis[camn]) +1)), repro_errs_analysis[camn])
+ ax.set_xlabel('x (m)')
+ ax.set_ylabel('y (m)')
+ plt.show()
+
+ return dlt_coefs, repro_errs
+
+
+def gen_dlt_coefs_from_paths(xyz_path: str, xy_path: str,
+ points_to_remove: List[int] = [], plot_error_analysis: bool = False,
+ from_matlab: bool = True, img_height: int = None) -> List[str]:
+ """
+ Generate DLT coefficients for multiple cameras from the paths of csv files with known 3d coordinates of the objects
+ and 2d coordinates for all objects for multiple cameras
+
+ Args:
+ xyz_path: Path of a csv file containing known 3d coordinates of the objects (size: nb_obj*3)
+ xy_path: Path of a csv file containing 2d coordinates for all objects for multiple cameras (size: nb_cam*nb_obj*3)
+ points_to_remove: List of indexes of points that need to be ignored for the computation of the DLT coefficients
+ plot_error_analysis: Whether or not will do an error analysis
+ (check if not using each 2d point will impact overall calibration error)
+
+ from_matlab: Whether or not to convert 2d coordinates to the coordinate system used here
+ img_height: Height of the images of all cameras (assume the same for all),
+ only needed when converting 2d coordinates from matlab
+
+ Returns:
+ dlt_coef: List of DLT coefficient (11*nb_cam) for all cameras
+ mean_repro_err: mean reprojection errors in pixel (or Root-mean-square deviation) for all cameras
+ """
+
+ coord_calib_csv = np.genfromtxt(xyz_path, delimiter=',', skip_header=0, names=True)
+ pts_calib_csv = np.genfromtxt(xy_path, delimiter=',', skip_header=0, names=True)
+
+ coord_calib = {'x': np.array([s for i, s in enumerate(coord_calib_csv['x']) if i+1 not in points_to_remove]),
+ 'y': np.array([s for i, s in enumerate(coord_calib_csv['y']) if i+1 not in points_to_remove]),
+ 'z': np.array([s for i, s in enumerate(coord_calib_csv['z']) if i+1 not in points_to_remove])}
+
+ nb_cam = int(len(pts_calib_csv[0])/2)
+
+ img_pts = [[]] * nb_cam
+ obj_coord = [[]] * nb_cam
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ x_pts = np.array([s for i, s in enumerate(pts_calib_csv['cam{0}_X'.format(camn)]) if i+1 not in points_to_remove])
+ y_pts = np.array([s for i, s in enumerate(pts_calib_csv['cam{0}_Y'.format(camn)]) if i+1 not in points_to_remove])
+
+ if from_matlab: # Will transform coordinate (from Matlab) system using img_size
+ if img_height is None:
+ raise ValueError('Need to set image height in order to convert coordinate system from matlab')
+
+ assert img_height >= np.nanmax(y_pts)
+
+ img_pts[i] = np.array([x_pts, img_height - y_pts], np.float32).T
+
+ else:
+ img_pts[i] = np.array([x_pts, y_pts], np.float32).T
+
+ index = ~np.isnan(img_pts[i]).any(axis=1) # Remove nan
+ img_pts[i] = img_pts[i][index]
+
+ obj_coord[i] = np.array([coord_calib['x'][index], coord_calib['y'][index], coord_calib['z'][index]]).T
+
+ return gen_dlt_coefs(obj_coord, img_pts, plot_error_analysis=plot_error_analysis)
+
diff --git a/data/calib/examples/mosquito_escape/20200306_xypts.csv b/data/calib/examples/mosquito_escape/20200306_xypts.csv
deleted file mode 100644
index 014a4b5eb1c2b441036cb31720c74994cb0b8cfd..0000000000000000000000000000000000000000
--- a/data/calib/examples/mosquito_escape/20200306_xypts.csv
+++ /dev/null
@@ -1,24 +0,0 @@
-cam1_X,cam1_Y,cam2_X,cam2_Y,cam3_X,cam3_Y
-nan,nan,nan,nan,nan,nan
-nan,nan,nan,nan,102,82
-nan,nan,nan,nan,438,526
-nan,nan,nan,nan,nan,nan
-53,752,nan,nan,nan,nan
-232,280,nan,nan,nan,nan
-510,838,nan,nan,1005,860
-758,610,60,272,938,626
-nan,nan,nan,nan,nan,nan
-nan,nan,nan,nan,986,926
-nan,nan,nan,nan,621,362
-nan,nan,nan,nan,nan,nan
-625,816,nan,nan,nan,nan
-424,378,901,626,157,398
-nan,nan,668,849,nan,nan
-273,633,406,781,590,634
-560,109,189,497,784,121
-nan,nan,329,166,nan,nan
-973,218,645,56,378,218
-nan,nan,829,61,160,1014
-822,81,nan,nan,10,114
-nan,nan,813,396,nan,nan
-530,506,514,521,498,512
diff --git a/data/calib/examples/mosquito_escapes/20200215_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/20200215_DLTcoefs-py.csv
new file mode 100644
index 0000000000000000000000000000000000000000..01959d6bb5290367ca3d3c4fdcfae194879d0205
--- /dev/null
+++ b/data/calib/examples/mosquito_escapes/20200215_DLTcoefs-py.csv
@@ -0,0 +1,12 @@
+# cam1,cam2,cam3
+-8.262424931591261839e+02,9.535763357450576223e+03,-6.658637507848138739e+02
+9.342390484220992221e+03,-1.984440135562464402e-01,9.057492290578740722e+03
+8.325857030634652745e+01,-7.371305819268055757e+02,5.808225753247376133e+02
+4.249044568903008781e+02,5.215101083947636198e+02,4.523248219467138824e+02
+-6.867280605283698378e+02,3.447013951805126197e+01,-6.963120258685201406e+03
+8.485001368738440419e+01,-9.541215128903701043e+03,-1.041321374154470725e+01
+-9.354011898208420462e+03,-6.532590573919845838e+02,-5.877425922192175676e+03
+5.378205057715513249e+02,6.051492814375513944e+02,5.088074918326084344e+02
+-1.494755255760544133e+00,5.309336695878297263e-03,-1.178273065522193352e+00
+-1.958966279489277040e-02,-5.772301911740258806e-02,-2.172570887228763518e-03
+1.168411603741917847e-02,-1.494943508532696530e+00,1.080794810979094134e+00
diff --git a/data/calib/examples/mosquito_escapes/20200306_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/20200306_DLTcoefs-py.csv
new file mode 100644
index 0000000000000000000000000000000000000000..d4d6c617ec8a36ab5ca50741fb55763f85ef1fa9
--- /dev/null
+++ b/data/calib/examples/mosquito_escapes/20200306_DLTcoefs-py.csv
@@ -0,0 +1,12 @@
+# cam1,cam2,cam3
+-1.141261605371704945e+03,1.328673407525115726e+04,-1.191831435894136303e+04
+1.309940896417702243e+04,-8.778595600402001153e+00,-1.019424781667450475e+03
+6.738506807383616604e+01,-8.981925484222546174e+02,-2.997667235750116177e+01
+3.842798595065022482e+02,5.319070289962245397e+02,4.773508932664243503e+02
+-1.116010261322133829e+03,-8.034049187175665452e+01,2.938439556082694253e+02
+-1.992891489952654638e+01,1.318195592211345684e+04,-1.005860989597624894e+03
+1.296662062858988247e+04,-1.082397032455340195e+03,1.188050139878365735e+04
+4.698677169860810636e+02,3.559214627515071925e+02,4.653246378138107957e+02
+-1.984822484276100418e+00,-6.581836073099980866e-02,2.467466773754960763e-01
+1.491439211012067223e-01,-1.480609988735411831e-01,-2.038602110252764277e+00
+-1.617512824958341477e-01,-1.859361725271246879e+00,-2.534475051759991215e-01
diff --git a/data/calib/20200306_xypts.csv b/data/calib/examples/mosquito_escapes/20200306_xypts.csv
similarity index 100%
rename from data/calib/20200306_xypts.csv
rename to data/calib/examples/mosquito_escapes/20200306_xypts.csv
diff --git a/data/calib/examples/mosquito_escapes/to_delete/20200215_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/to_delete/20200215_DLTcoefs-py.csv
new file mode 100644
index 0000000000000000000000000000000000000000..01959d6bb5290367ca3d3c4fdcfae194879d0205
--- /dev/null
+++ b/data/calib/examples/mosquito_escapes/to_delete/20200215_DLTcoefs-py.csv
@@ -0,0 +1,12 @@
+# cam1,cam2,cam3
+-8.262424931591261839e+02,9.535763357450576223e+03,-6.658637507848138739e+02
+9.342390484220992221e+03,-1.984440135562464402e-01,9.057492290578740722e+03
+8.325857030634652745e+01,-7.371305819268055757e+02,5.808225753247376133e+02
+4.249044568903008781e+02,5.215101083947636198e+02,4.523248219467138824e+02
+-6.867280605283698378e+02,3.447013951805126197e+01,-6.963120258685201406e+03
+8.485001368738440419e+01,-9.541215128903701043e+03,-1.041321374154470725e+01
+-9.354011898208420462e+03,-6.532590573919845838e+02,-5.877425922192175676e+03
+5.378205057715513249e+02,6.051492814375513944e+02,5.088074918326084344e+02
+-1.494755255760544133e+00,5.309336695878297263e-03,-1.178273065522193352e+00
+-1.958966279489277040e-02,-5.772301911740258806e-02,-2.172570887228763518e-03
+1.168411603741917847e-02,-1.494943508532696530e+00,1.080794810979094134e+00
diff --git a/data/calib/examples/mosquito_escapes/to_delete/20200215_DLTcoefs.csv b/data/calib/examples/mosquito_escapes/to_delete/20200215_DLTcoefs.csv
new file mode 100644
index 0000000000000000000000000000000000000000..9511de80799a51971e6282067042ab6a81564ab5
--- /dev/null
+++ b/data/calib/examples/mosquito_escapes/to_delete/20200215_DLTcoefs.csv
@@ -0,0 +1,11 @@
+-805.02,9557.8,-647.01
+9352.5,7.9159,9044
+74.967,-740.34,591.92
+424.71,522.15,452.16
+-791.52,-35.483,5810.1
+-95.262,9509.3,-13.818
+9352.5,-855.76,6980.8
+485.57,419.18,515.86
+-1.4474,-0.017694,-1.1411
+-0.0088429,-0.015916,-0.026959
+-0.014362,-1.4501,1.1051
diff --git a/data/calib/20200305_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/to_delete/20200305_DLTcoefs-py.csv
similarity index 100%
rename from data/calib/20200305_DLTcoefs-py.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200305_DLTcoefs-py.csv
diff --git a/data/calib/20200306_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/to_delete/20200306_DLTcoefs-py.csv
similarity index 100%
rename from data/calib/20200306_DLTcoefs-py.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200306_DLTcoefs-py.csv
diff --git a/data/calib/20200311_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/to_delete/20200311_DLTcoefs-py.csv
similarity index 100%
rename from data/calib/20200311_DLTcoefs-py.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200311_DLTcoefs-py.csv
diff --git a/data/calib/20200311_xypts.csv b/data/calib/examples/mosquito_escapes/to_delete/20200311_xypts.csv
similarity index 100%
rename from data/calib/20200311_xypts.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200311_xypts.csv
diff --git a/data/calib/20200401_1531_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/to_delete/20200401_1531_DLTcoefs-py.csv
similarity index 100%
rename from data/calib/20200401_1531_DLTcoefs-py.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200401_1531_DLTcoefs-py.csv
diff --git a/data/calib/20200401_1531_xypts.csv b/data/calib/examples/mosquito_escapes/to_delete/20200401_1531_xypts.csv
similarity index 100%
rename from data/calib/20200401_1531_xypts.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200401_1531_xypts.csv
diff --git a/data/calib/20200615_DLTcoefs-py.csv b/data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs-py.csv
similarity index 100%
rename from data/calib/20200615_DLTcoefs-py.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs-py.csv
diff --git a/data/calib/20200615_DLTcoefs.csv b/data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs.csv
similarity index 100%
rename from data/calib/20200615_DLTcoefs.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs.csv
diff --git a/data/calib/20200615_xypts-rot.csv b/data/calib/examples/mosquito_escapes/to_delete/20200615_xypts-rot.csv
similarity index 100%
rename from data/calib/20200615_xypts-rot.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200615_xypts-rot.csv
diff --git a/data/calib/20200615_xypts.csv b/data/calib/examples/mosquito_escapes/to_delete/20200615_xypts.csv
similarity index 100%
rename from data/calib/20200615_xypts.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200615_xypts.csv
diff --git a/data/calib/20200623_DLTcoefs.csv b/data/calib/examples/mosquito_escapes/to_delete/20200623_DLTcoefs.csv
similarity index 100%
rename from data/calib/20200623_DLTcoefs.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200623_DLTcoefs.csv
diff --git a/data/calib/20200623_xypts-rot.csv b/data/calib/examples/mosquito_escapes/to_delete/20200623_xypts-rot.csv
similarity index 100%
rename from data/calib/20200623_xypts-rot.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200623_xypts-rot.csv
diff --git a/data/calib/20200623_xypts.csv b/data/calib/examples/mosquito_escapes/to_delete/20200623_xypts.csv
similarity index 100%
rename from data/calib/20200623_xypts.csv
rename to data/calib/examples/mosquito_escapes/to_delete/20200623_xypts.csv
diff --git a/data/calib/xyz_calibration_device_big.csv b/data/calib/examples/mosquito_escapes/to_delete/xyz_calibration_device_big.csv
similarity index 100%
rename from data/calib/xyz_calibration_device_big.csv
rename to data/calib/examples/mosquito_escapes/to_delete/xyz_calibration_device_big.csv
diff --git a/data/calib/xyz_calibration_device_small.csv b/data/calib/examples/mosquito_escapes/to_delete/xyz_calibration_device_small.csv
similarity index 100%
rename from data/calib/xyz_calibration_device_small.csv
rename to data/calib/examples/mosquito_escapes/to_delete/xyz_calibration_device_small.csv
diff --git a/data/calib/examples/mosquito_escape/xyz_calibration_device.csv b/data/calib/examples/mosquito_escapes/xyz_calibration_device.csv
similarity index 100%
rename from data/calib/examples/mosquito_escape/xyz_calibration_device.csv
rename to data/calib/examples/mosquito_escapes/xyz_calibration_device.csv
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010001.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8540072cabeb62484866c8277c5edc542b6c8ef5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010051.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ca26986d1730cb5f9e4c91fc572e83b1a1a6e171
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010101.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..40dece84a99917361625149bce1e2e2dda52939c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010151.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cc52bb1b7e0a6bd6e8bf17be6ece2e0e4b3d259b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010201.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c790644b27cfc656d175f745ceacaae8b50f9c7e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010251.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..366fea0f5550a7e9b529c22db636bd96dc155622
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010301.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5e70b2f8fee48a3dfeb2624d58d017fdf1d0a9a5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010351.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f0063e3707a8c6ad8d5fb0e41c1bd2494c2f3964
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010401.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6f321f3909be553bef3c04f804cab4711d3259a3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010451.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..38c3cf12267953095830ab8c88af9243bc32833d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010501.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..915309d1afaba3245a38a6bf62865229d61b30af
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010551.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5cee961a2f03e2592b3eb3b980251991b0df003e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010601.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b89d1a7a3034ea1d498edac720edf3d722b949be
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010651.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0fc176212ae3c76ad5aa8a348c0ed5a19a14fa3f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010701.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..54e1eb1b314f37886cf221fc9499c517bf898edf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010751.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6b788f3c72013d96048da9dcaa1ef056699872ac
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010801.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6e46acb6c58d3a623686e5de218120b730a4cec0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010851.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7b35b1dc3fb64a69a079faa4158e1d18c9573abe
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010901.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5b126dd0083c61f6fbab36da98e48b4c984c30ee
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010951.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b799d94e72b303345774bea842008ba9e32d417e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759010951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011001.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..55263b4dfae45415026910f2e553d9142509cc83
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011051.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4d58967b1bf3d576ce2f37a939a976fce5ce1c27
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011101.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3f4e9484d46ba91f22ebb369d77aa15b4fd52fb9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011151.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..de2dd0bafb8fe55e4c99228a9aacd3819d79d38f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011201.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..955cc9d0d793b608a58f5107727d69d840050567
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011251.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..614bb3fe81fc77f60018eab387c1f19ffc6a388b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011301.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..82940ae16884df0ed76ee36a047b57b393d05945
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011351.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f803995e7f77c2b54e475959b31c35b8e6dd1c0b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011401.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0e078fd95c2f37fa486f4e434d8c7d5af461a7c2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011451.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9e867f54cdabf0e3390c5f5c65b7ba6c5b047ea4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011501.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c934e8c8fdf4db864054cc2ebf9b6098a6e7dc4e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011551.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1bef9288880447ac1f73ca8e8ded5ccd96874cf8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011601.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4082cc7affc677dd1d7a3a12cb1292f6fb452b71
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011651.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..aaf5c500508598adf8e1c0ac8761486ce6d00d37
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011701.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..22531340927812f7c97948e05c33dfd890d6e9eb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011751.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f752ae296ed525e25abb37fc118cb85c9f4b334f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011801.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..71d2819e90a7ba1b00581127d1066ced9388696f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011851.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7e14f9e2447ec11a5f8677ae27769d3e65030916
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011901.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bbec2966ba8f0707553572c9b91a673563778d9a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011951.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6b2b3dafa122eb025d0f593f7ad07b71176fe48d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759011951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012001.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7eac4df44dc6e676d00aed2e8976c4107be58c02
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012051.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..59be7b352da90de5edaeedcc1e6d332cd20d1319
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012101.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8140606cd718a225c1be852a6aa6e7701b09491b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012151.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e650ecc44c04b415c8eb180d9384ec313ce3e6f5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012201.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4c1f2f8747d4973828a1689790c10cae6a420178
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012251.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..13711d48bc95140534b67f02021916089f71d0a4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012301.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4b7e10459a408a0cd19333fee269fca078c86549
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012351.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..12a5bcd893d25af62d378658737667aecc7aca8c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012401.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3e20c382d691940ae51f3a823f1574c7d1e9f1b6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012451.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..faf33b0f77ed50a5607bd227ad47ff893ce53e90
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012501.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0e1bd68c03252a673e16b0d0c282e6f567b83fd0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012551.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..52229f12f079bed5f8d78dcc52bf6a2befaf5af1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012601.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a5caf7a4dd8805dd3a29377d25aedd5b4060d5a9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012651.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dbf89b0b3db7a41df6cc0ee6803e9a4d6f23fdd3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012701.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c6b42841a769ca27639245ef6763b42432e910cb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012751.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6615753827421571013455261a63ca436ae412cf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012801.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..14eafe8617124ec2cf7309cc0eb005d3d0275993
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012851.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6a99333aab2c6e808a1c60b7aae7fefd9d753265
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012901.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3a43bf571d8af6c3b8a797821d8aeee4497f2b2c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012951.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8ffb8457038e7fba93bff4ccb9fcda25a43f8f06
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759012951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013001.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7843af1d10ade00e95ad26d283b6732f7669ea60
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013051.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b8a532f2f6e25f6976128d8c084c5b2110d7702d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013101.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4b5e5058a1bf48ae91b24aaaa7dc9338a6dc4624
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013151.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..12cde6f365c98c0b185d7833213cf668700c992e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013201.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d908f499ceee08fdb96b102d47465773fe611917
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013251.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..82a1ed51f77b4292a537250069d90cf3eb49401d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013301.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..660a6492f683f89a2bea993cfc3b1bc76418ced0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013351.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5d39ecf7322115c8006da569c56bc44fc0baf41c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013401.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f8832919819ba154ec960ac0f3a14136cef63fbc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013451.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..440dc2fc34c9b162654c2a53c5a7f3894a2c885a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013501.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9a35e1ce36999c19f098317b3488d8954e093706
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013551.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b6856444bc93f5fd66d4b5982e7e6817522e2796
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013601.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b8fe5b3e4d034f48870a55b23c33286c6f92a5c2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013651.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..496520b9275095c78a57209604224ded5cd830f8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013701.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..91d44d2cd9cd363bbbfbf35a8e0b24f2c9051c7a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013751.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c1cad8841010a3ee6a0efe96ed91e65196edbb33
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013801.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8d6774467e468cb89aafea59718d9cb40618ef35
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013851.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..45af8bdb5c97da1e92ba220796a1ce75e9370061
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013901.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2511d01730c6d53fcf00cd4f0c78e27601d610e3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013951.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3f1e97072d9529782d6310be82e024013b96b6b4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759013951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014001.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..572cbb9ffa92e650d9bc4b3067d0692f4c40503f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014051.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1c72aa60105869290db293cb596106d5ef644ffb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014101.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..52818872dafcbc27f941e76913dc285efbed448a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014151.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1fb78dcbc2f1f91c3562e9764efaec14eb723d3b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014201.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..92b3e083ca9eb852186840738d8788d8358f040d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014251.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e5475d9d785cbb6d7737de507c2f17a9a85ff7b7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014301.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7e5a84bc70c0a97285a9025c6e3a7335b62e3609
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014351.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..336d7b0ab1177eb90a7b438f7572e1efbc224862
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014401.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..07f15fb414d070e5f84ebc226b7e928e8866453b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014451.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f34dd175af6f6f2a307e7fce90aca65d298f4d2b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014501.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..17e4250afd8cd115528df8e539d03b6b4c18e8c3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014551.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..71c531198f74c2e9355383b21c5c4beb47230ce5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014601.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2e189c67a5b2954fd09c5b07056c26991fab2399
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014651.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c7b48d0ba248eada5f1a14c73fec04e96c050840
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014701.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..475abf063e55b6a1157cdef1983f15351ea69a47
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014751.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3ad6f3cc0fba858faf54bba23976e7d64811840c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014801.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bc4b1b55369eb5785d9a6dee6b80a91dbcdc0b42
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014851.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6f2bec8b27864f18c0b8ff2b91bf9c4742b54b21
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014901.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7bf8a4c8afa2341712c40b5f15500d715deefa7a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014951.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..991074c180ed06229138a427736784e71a3f4e42
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759014951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015001.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..742b39e14db9db72df3056c8c1bff357677a61f8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015051.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..adb14a64cdd3b8c19e93a784aa8328676dadeff2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015101.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2345628c3edba8241d0c51e0c8f4e356584e52c6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015151.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fdbfa7bee4a5f0b62f6c75d68d72304bd74df457
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015201.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..09d572eb55dd958be9eb52aea477ad718676f693
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015251.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fb690fd43d8df6dacef2cbbbbba6448e3dc87917
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015301.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..31dcb1cec245806dfe3486e760dc532f9a0a018d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015351.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d07d56f5c11a30ac2319b87ffd2e1d50cc370906
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015401.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d08f867283d179b6b8c825fd32e09498e40a8be1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015451.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..786fdc777183da37d86e92f9cb3e95bec650b702
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-sample_binned/cam1_20200216_0759015451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013501.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5ccb573c9d1e02d185a0cf3150dddde8d2fb9194
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013502.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013502.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5eea156a690ff2fc173762ee94d1fe248d0023d0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013502.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013503.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013503.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0fa0a1d04cac18607751f54e233985aa9c27d3e2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013503.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013504.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013504.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5c2847aadb0e6853fdd7c3aefd197253f90c6cd4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013504.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013505.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013505.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3650c53d5348755704e19c5367923e94b0b816fb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013505.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013506.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013506.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..44d9442f5ce9e3b36db2d818398141f0c6e0fe16
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013506.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013507.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013507.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..84a85c0359cb7acf5ed3b8bfba20c58fc009edc3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013507.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013508.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013508.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..21b081f54967b3bf1abf5659537d5aad0ded95a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013508.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013509.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013509.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9ab4ca7a385acbcf25d4fb1f7209b4c30bf03732
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013509.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013510.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013510.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f6f9a0324a396e689977510f5a25b7665ce43f9a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013510.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013511.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013511.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dae1a8d516a1d4f6f8044b8c3267d04fcabd7645
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013511.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013512.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013512.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c70f059561dfce5dd19200f946efefb2776da448
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013512.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013513.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013513.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5591a203d832f153fb4f70282e99126bfb926f2a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013513.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013514.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013514.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6f079e230be201f0333b5d984c044662669bfe27
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013514.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013515.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013515.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..60fe7164828d2786feee52066310c41ea42d22a6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013515.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013516.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013516.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b786156a0d0d7f1f68dac5ee5b76ddecaef42a87
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013516.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013517.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013517.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3389cd69804cf0180c170a683571023427e0b081
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013517.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013518.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013518.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4e79e03725fb8a393267f8e9629fb2e6f290fb0b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013518.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013519.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013519.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..446886f0d650ebbd4792d04dcd863efea4278ecc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013519.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013520.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013520.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1effd4da987913a562bf64551dab236397e35a00
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013520.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013521.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013521.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..34948eb4858506befbf10f0edc42d10acd46303f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013521.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013522.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013522.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..55ef3d44fc634bd52951f94ce76e6ced606be92f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013522.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013523.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013523.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b72cca997538c6379c32ef07cc831c30dd267557
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013523.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013524.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013524.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2a4e9787f608738256c470ef542febd3c47ee1d3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013524.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013525.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013525.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2f200a60aa6d5852f5d37b1c8f3c1fe9598d913f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013525.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013526.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013526.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3d37a98bae0a671c48b6e7baae094163db280c81
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013526.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013527.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013527.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..98716f257753cdfb7a645e83443331505bf5c725
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013527.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013528.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013528.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0c563270bc3b0459d833399e0bef6ad960d56739
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013528.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013529.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013529.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..65531eeb4d1288ac750b3ae8c623117cbc5c68e7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013529.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013530.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013530.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..02d3c3c25ba195c65556c11dc201d24d1e680e8c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013530.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013531.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013531.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b03df6243c821980e9001159f7368e61554c3331
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013531.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013532.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013532.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3ca7f1c6a362fdcf59bdfdd9a30f1e33bb502e1c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013532.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013533.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013533.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3fdc0cab6df67f279e5db97cd190ff7e1ba52fae
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013533.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013534.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013534.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..341e2f1b38a7367968449f04e054f7c2803c9c36
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013534.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013535.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013535.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d7dec3921d9e7759dd18bb71f101e4b5a9afe8ce
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013535.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013536.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013536.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f65e4dfed0e1d9be2b6c38dff7314f121ca51826
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013536.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013537.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013537.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..087eb947170932faaf6b1ec7a3d13cf932eec27f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013537.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013538.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013538.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bb75901b805eae9d6dd6149dbd803b7b80df9f29
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013538.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013539.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013539.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5b855ba8e76c66f0ca340d5435437983ad7537ff
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013539.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013540.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013540.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ebf82545fcbb9680c8702e3d866b2d84cd7f4c2c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013540.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013541.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013541.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..67b9d4d7c0e9b5429df0bab9bc8dd13c4e22b020
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013541.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013542.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013542.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fb8fb23684f92d2a1d3424d2b02af9ceb26a3d3c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013542.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013543.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013543.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..016b4595b2c49842fbb091136583fef32d33e10a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013543.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013544.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013544.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c1bf5f7cc9e31cff228fe3f9f400a13d05001706
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013544.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013545.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013545.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0355c03c565dbf41edb9af443286f073a8acfde9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013545.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013546.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013546.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cd12dab915a73891ccc4a607c2d0531a45754f50
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013546.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013547.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013547.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e693cc0ceb1e7adb3763a1fc70cff4b914ac71b4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013547.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013548.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013548.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cb4f18b89fcdd69c9da0e207c79c94fc27fbb02a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013548.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013549.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013549.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..147b448ba0d1421973df58fed3b485988bae9867
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013549.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013550.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013550.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..422d52d79ac5f82cf591f38785cde14bb09257c0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013550.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013551.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8e4b49d66d6d942813657f2a1f74685e8c785103
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013552.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013552.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..21b2b413605310fea83322f8198a810663cd33c0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013552.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013553.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013553.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f4826b70bde3dde6c05ca86973280cefa35eca27
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013553.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013554.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013554.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..574b545c826afce716ad9b581831597625197ecc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013554.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013555.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013555.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..773d23ebcbaad539ae5bf8e1a6394f09086448c7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013555.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013556.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013556.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..439df3ddd57adafa6a91acc4291401110d4be016
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013556.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013557.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013557.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0abaf0e69613dfabfe2e0906c09d6ddb3ce51520
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013557.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013558.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013558.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e94a2492fbefa36bbf82ba5b1cec4bec5ef412ba
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013558.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013559.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013559.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7bb732f78088895477daf1afc4402ceeaebdf7a4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013559.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013560.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013560.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..aa1646a366f79337333d165154288196d0bf3a4e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013560.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013561.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013561.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..301373578edc82e79039b566c3547cb9052d7795
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013561.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013562.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013562.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e6428fec3370eee40c35a61a21c9fc18318f3b6a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013562.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013563.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013563.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..192cff7383eaeaebf925a307d7d4138ccb62804d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013563.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013564.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013564.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6e281777667c7cd3dd99edfb37708b2245b8a44a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013564.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013565.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013565.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..02abfe3c8e451276f2af1bf419adec624ac62483
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013565.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013566.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013566.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..82c34fe7aba22f0fa05ea74e9d5443bcee9ef5ca
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013566.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013567.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013567.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..96796fe824cb98415c3065699f5ef83864de15d2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013567.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013568.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013568.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c124c8aea95b09dd154d30d3442bce2ba1767111
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013568.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013569.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013569.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2e14e022cd9d52a6758a36c433fb48a4a2a0add1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013569.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013570.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013570.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..048758d1b71827e31cf9cfde28299ab08ee185d0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013570.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013571.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013571.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a4b06043e34cd9a72c37dec8af31e94328ef95b0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013571.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013572.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013572.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c6510a4222506d424ce5e5a794742dcd15f6ca0b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013572.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013573.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013573.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..46dd2c53b11966bd379794ba7d64b887a956ca9e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013573.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013574.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013574.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..db8332724e21c7db6b26cf081616cfe2a9f2b7f0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013574.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013575.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013575.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c3511614e461d7e0125645179329e22f15789f42
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013575.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013576.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013576.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1289e6ec2e035524c22c25bb7cbdd35063ffa7e6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013576.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013577.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013577.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7ecb9fc63ae79579869dbcc11badcb95f762d352
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013577.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013578.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013578.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1d27077b8a2d46fe79a2dc492ed656f8a1008cc2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013578.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013579.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013579.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e82f1fadd4ffcab78b1d043d0d38f969e324e344
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013579.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013580.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013580.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b1180f0dd9a9e9009932e9a97bc28a1b3af5c09e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013580.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013581.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013581.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0d6097cb900c1b6a5f4e84cee808ba55e217abde
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013581.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013582.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013582.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..38f57d3d1ac6ae7ebc32189081fd26f42ba2fba9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013582.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013583.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013583.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a7b48990a0722291599de80936f6b41a7e2913b4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013583.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013584.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013584.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..337ed896c63f380cc50b4ea490fd44ca0a9f3dc2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013584.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013585.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013585.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9cb5e7d9213d9907422942907ec4900066361726
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013585.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013586.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013586.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d29bf4fffa7e39b62e126b6966d06391180907a0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013586.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013587.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013587.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5e8069dbdd7cdcf60cca80a1c4731becdd71beef
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013587.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013588.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013588.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..043ca63748edfb302db4ed2dd1bc3aae0c027307
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013588.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013589.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013589.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4f3aa2351ae5d43255b698a8a492a37a12ddd48e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013589.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013590.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013590.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6ec3949947f02cc8d399e880f441faf7104dea45
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013590.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013591.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013591.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d91d4921bdbf7234ef7fb0ed65a45352ad7cb954
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013591.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013592.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013592.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cc77ff93f18d327d96af7b27e90d88a746669346
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013592.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013593.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013593.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d855478dcb2d76ea45c2ab9e3559e240a6d07430
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013593.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013594.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013594.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2f7db7b5b253cf0bb764e1cd2104a420ff1c2157
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013594.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013595.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013595.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..59b517efdacb48361bcfacbee836c52b7c2a7507
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013595.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013596.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013596.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8cef1527a2e8d474bc91d366b2035f84922c728a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013596.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013597.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013597.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b1c15dfcf36379d97f822545ddbf72880a360335
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013597.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013598.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013598.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cdadb815023e6fb7c2da617418405ab5455d7aa0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013598.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013599.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013599.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c8003b06ca6e5ab97f5c0546bb51da4214fb3722
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013599.jpg differ
diff --git a/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013600.jpg b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013600.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a030a59bc04c3db8426f1202e648321e5b4def1e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013600.jpg differ
diff --git a/data/examples/mosquito_escape/cam1_20200303_030117/cam1_20200303_030117-2d_points.csv b/data/examples/mosquito_escapes/cam1/20200303_030117/cam1_20200303_030117-2d_points.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam1_20200303_030117/cam1_20200303_030117-2d_points.csv
rename to data/examples/mosquito_escapes/cam1/20200303_030117/cam1_20200303_030117-2d_points.csv
diff --git a/data/examples/mosquito_escape/cam1_20200303_030117/cam1_20200303_030117-obj1-2d_points.csv b/data/examples/mosquito_escapes/cam1/20200303_030117/cam1_20200303_030117-obj1-2d_points.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam1_20200303_030117/cam1_20200303_030117-obj1-2d_points.csv
rename to data/examples/mosquito_escapes/cam1/20200303_030117/cam1_20200303_030117-obj1-2d_points.csv
diff --git a/data/examples/mosquito_escape/cam1_20200303_030117/cam1_20200303_030117-obj1-3d_track.csv b/data/examples/mosquito_escapes/cam1/20200303_030117/cam1_20200303_030117-obj1-3d_track.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam1_20200303_030117/cam1_20200303_030117-obj1-3d_track.csv
rename to data/examples/mosquito_escapes/cam1/20200303_030117/cam1_20200303_030117-obj1-3d_track.csv
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050001.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5ddcb8bfa18e37b4d108461dc86751041c3275b6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050051.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b7c7ed1a628b59de179f4058489b73b20955f874
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050101.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e8f34e6f1d9e8adb416f357a8b010f337e9301d6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050151.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ead87ef1d7dddbf0d80ba2161bd105c696d635c8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050201.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..af2e77ee6768577a66164fabf8895cfe52df5e9e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050251.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0513d1f57a4374da60fe8092313472a876ba146c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050301.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c5d9644c426ce9b4d0b42b35621f8f399a5b84fd
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050351.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..95e3dc4314bbd04a9277da1059399d242f5a7292
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050401.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d2e1b36a5f0fd19100e2e07a4c0fe2260616ac43
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050451.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0ed8bfc97579a28d6407236dfaec5e864ac6ae9a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050501.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f5a859d3f083dc659c45488f846c461c00c093aa
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050551.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0a12cf1fe62e203e935ce315a7da668128adda11
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050601.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..db44efa1738a0118447afc332c5715d9d3065ba9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050651.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..79d8c4944fa5946aefd3c524e3caf7567694cb76
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050701.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..902732982b90cd90d2eb4127ce5cacc7a08aa758
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050751.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4766e05bd8d18c8038a6b6cc888e6b2e23ec221c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050801.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5a34b8e5d1a791198880b5d553b1118195000357
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050851.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cb4438c418d91b4bff64742a6aade057afedb49b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050901.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..49b263494429479732247cb47d5559b91318c147
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050951.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..53d5a89791f4a2db6ac7c926a0273dc9a19e2a48
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759050951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051001.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f15ca701366d5fcbbcb538f172c80dd7d66fa4a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051051.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d6937c290f6305fdc6d6adb91dd7b0e39b04e619
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051101.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..66f3f02023977b948ad16b358d679f48fded4685
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051151.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..31203dbc715ae421c5813a1dd3356c828adeb7fb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051201.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..10c87a301efaf2bea0b7afea723ccf2cae373c7e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051251.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8131dbc12b88df01504cb452249b6c5c7eafa3e9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051301.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..27a6b815cea811b4dd7e87c4f8f4bd8734ce1851
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051351.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2b68ef2e26144dbe24672644f469cd3c05641633
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051401.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9dbf5c9ab71e56008a8ea8fe56068afa2714b949
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051451.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fb231b4e83424267afe828e9ef4fcfe892b2403a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051501.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ca339807326ffec23a3a57779ca3cfd9398322d2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051551.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ba09ba77730020f6968287569ceba9f5d0aeb9d3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051601.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7c8d2ab755ccd6b2acbc6aeff7f37b3aaf93a39b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051651.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c8e5e7a0e903ccea32cddc89481249065f06220b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051701.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f575e4bd9d6e9c130952fd78fb97b91925af5e46
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051751.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c6523557b6fcb605f86e886b722147cb59e65655
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051801.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cd8bdcffda543009c1f65aedebd2eb630b67d014
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051851.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ea60d40e07a9572284aac546e121aeeb4c15a374
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051901.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a5e106eee95109b85dccc18033f870f6ba9c5bce
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051951.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7ad2063d31ee3258b9834322ded4af8c5db7056b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759051951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052001.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..341643a30bad8bfca60c98f15c3f7f5fcd5f364a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052051.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..15a72dd823b439d4cb0d2133f00b406d2744f2f6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052101.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1a162e355720f7ef7eb1383e4b608cebaba9730b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052151.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b9683c0e8363625d3b1ebf05303135610e156820
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052201.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bc5ab73f14fc8332423e4bfe98427b9de7cc4016
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052251.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4def47763251f7a1293af53b8fe3ebeae872d795
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052301.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d3581c3385c5c6e1b460b07e83d465ef59042206
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052351.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..46d5bc81f9c47c60c3b04d6cbcf822e94becce83
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052401.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..49b492a5e9b61de5ea926837d066fe532abfca6a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052451.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7f2b72ff25637dcea67d231734a3bfc3057181ab
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052501.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8166f59d8455a50d83fc0e162b31069bafcc22c6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052551.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fcb53ee5106bc43ace6f483ccbab1f779e54eec7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052601.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..11b6db1771363182ff507cbb28225e8cfa82a810
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052651.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9f7191effd010093cf53148e680666dfe36350a4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052701.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..16341d14a6b8b089498f247118cacdd3c104d6fd
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052751.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..456bd4586d1dc018d34aa1b913efec5913961a14
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052801.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..53d1e314dbdf6f35478d502ab0ee08ee31a6b3c5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052851.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b9ffda5730e3d9849806bb4a34e437cae05485f4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052901.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8bbf9c051c51f8f4270e923e48bc8424f8f9d63b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052951.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..06eb3ec1ab1515f7526d8395e2364f776c67a67f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759052951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053001.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6f49364ab77d193ad125210817bfbe77e2e4a44e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053051.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d82006351242219f6e3a112f8e9f6346ff16ba06
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053101.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ea5985cb3feeaf399db56088e1d6b80d6a06af66
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053151.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cbe87dc576271ea5966248e087e607ba945edfc1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053201.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..340b3d3ad9aaed7728804b743ac6c467117715b0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053251.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..37cefc8b53fa87547f6e281c871688d902e415df
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053301.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..73a00cc2afbe180e6f52751b9d214724e87317d6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053351.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9fab5aecb643712a2e539ae7314c01cdb0c085d8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053401.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..991567530b1c71ede4bdb62f0a2d5a8df7c3a76a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053451.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0be1980be3af15923f809db8b101a63dfcc6d8ea
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053501.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2aefe9f70be10d6f73f914b52ec64b291addf82c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053551.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..aa3862065b753aaf8e84d78e652eae7a5910f99e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053601.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bc8dd050dace59b9efd5c05f6e796325e06fe0c8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053651.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d138a39fcf5babc4c3219cb63d9020dd75cf9729
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053701.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4fecbb146ef80a11f50da1cc625926f25acc6891
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053751.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..94da7603388a6732336ef48de63adf971c8cf06b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053801.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d450b4eb48c7eb9f9ca87aca2232b83401fa01d7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053851.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c6d5882976af342676b4b2e013a4ccc512db3a9c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053901.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..22d0b133f912992b422a94d6fbd8ab6919393139
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053951.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..63b1a5662481d2302c0b8249b7b9817acfcdcd52
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759053951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054001.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c7a254470d914ddc43bc5d1eb5254a8e197801f9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054051.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c71663eafe1a2b23a42125bbe5a99f2ef4721fae
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054101.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1578e99a39359345c9e9c6ecb89129663df69d5f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054151.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f525aa43f0409482f9749f3142457137b4406a30
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054201.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6fbe0d01c50465077c2c91c7d501e5cd62fe457a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054251.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..43c106a7197a8a5c765de66f12433371b1321809
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054301.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..69f7eacdb5c62a4ec567a180aa7231e0e93a0cbc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054351.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4b74c1e90500d50d186e5a3bfa1f474452bfbc3f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054401.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..df5ee580d84ae28f5dc3a45cfad10ec7381e07ee
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054451.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fa7b3936474e043e51c1b6a185440e137f1f3a18
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054501.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fcbd7bef6cb3ac8b82858effcdfb4cd9ae897094
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054551.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..982e6a9de5d8bbfce98ff807088d1d4393ef31c8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054601.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4120cebeccc6a3a3f5f7d652b9e06e12749190c1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054651.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0dbfc4a2468ba21a1b28c93b719f6f58cdf7da47
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054701.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..03d7b110ceff32519977f3eaab5d7ea00d68967d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054751.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..918118f9ae790e840789f4d5022fe86264057611
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054801.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e3c3ee340a5ed33a11cdb6971bf5ee3413042974
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054851.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..668a8af280d4de42378c9f082a166bee6ad90cfe
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054901.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3daff121200864cd7386ef9758e6832b59adee35
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054951.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..38d6d472e2343b0fa539fa2cc8dd7aab2c25043c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759054951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055001.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3b64103d9e24f4de0dca580a3697add5ece7c633
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055051.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..358f5d234a8ad2a75e03de1c644851ddba157dcd
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055101.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c4bf4c120579c00f65b743e1cbe639c231b50ab9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055151.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c828a36b5fa63d1636bc9cf1abff17b2892d85c7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055201.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1ed46669094fe5b5cf8f499757426002a30e4c7c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055251.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..59747f5b1b82fe69a7b7436aee2cae95ea7955c4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055301.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2313893551e1229045865af183be8a3857874941
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055351.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8e6211452edb9b87371b67d00adcdf18b4fb4a42
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055401.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..193825f4c1bb0e9dddd83776a7fc4b543db13269
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055451.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fc00d746c2e255ee58302d9efd5834987efc5e55
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-sample_binned/cam2_20200216_0759055451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053501.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c6ffda3e8c3ed1f6a02261992bf4d6eda64c3d5d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053502.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053502.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d0f41d85f753123c053010b627a278476f332774
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053502.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053503.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053503.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ea3260d82156e9584cca1d5bc6f022958b001a59
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053503.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053504.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053504.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3f6dba431c167de702161bfd3ca0d92272ab56a4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053504.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053505.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053505.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ee35347b632392dca2b1724394572b5b80775b34
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053505.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053506.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053506.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cc516c93e6b028e0645f34450a44f15d376b5a68
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053506.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053507.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053507.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3f97e17819a1972641a9e83ee11c2c55af2ab10b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053507.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053508.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053508.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ee911976a26b849e1fab47d2f64b501cdd88cc2a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053508.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053509.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053509.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..81e13041392ea3e5bcbce85cae25bad400ea69a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053509.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053510.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053510.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..07f07a8b3c0790aab975aa69bd1fdad0bb732daf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053510.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053511.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053511.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..074513c252331afb9d56f534cc1c6489ce29793f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053511.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053512.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053512.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4e85e0f4dbffd891eceed54cff75396811202eb9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053512.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053513.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053513.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..93fe2b5cad37a0e2442a343a99723a5b04d7de65
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053513.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053514.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053514.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..41894240ec1a8d51f20c0a4a0c875a7b5d1ae77d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053514.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053515.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053515.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..afdd7491222a5d616030ff89d26cd5cc5ba3c5d9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053515.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053516.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053516.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..117e9bb3466031d07954a79de6ea8437a5ca5248
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053516.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053517.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053517.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..40188df3cdd280bc76a74637f98d0b1876cdbd06
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053517.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053518.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053518.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2a8daffa797bad37062e97724fd632565cc61328
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053518.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053519.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053519.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..99ebf238072449c3b32297fe6cb50d8d22e8e875
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053519.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053520.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053520.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c1fc173d6ee2a6e3343b65710c11668ea91e9300
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053520.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053521.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053521.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..190ca6a15480274e56fe9232ec457df609495235
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053521.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053522.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053522.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4bdf29c2a610c89bb57befcc2ce497abac194d15
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053522.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053523.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053523.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..62f608b105fb7b3e4fc21ece16c62814c282b459
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053523.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053524.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053524.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9b11dd751b4176cac50325413c4a95b25c789b53
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053524.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053525.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053525.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1a765c281541a9da19e2f8f26bb2f74e792c7df5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053525.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053526.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053526.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6dcc8dc6f38cb6a677a719b40b9a46f31065c5cc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053526.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053527.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053527.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..caa4bdf7ae39ad42c8f3bd47a43702f86e67dd08
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053527.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053528.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053528.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..40a867ad90fd7c48090be8d719541bb5754d80f6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053528.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053529.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053529.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..424323080ecb746ec5dcf518f8802c21f7a0fd08
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053529.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053530.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053530.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5aa46504c7e53d13340c62caf12c11848ebcfb9e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053530.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053531.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053531.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..47e66dddb49dc5ed434200abc9e626fde98e5279
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053531.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053532.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053532.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d29e19730728e9f7a24fc447734dddc961c16caf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053532.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053533.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053533.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b4df1dc5ef007d4e4a5109790505ac5c5e523265
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053533.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053534.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053534.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..878597506bbcd40d86a2ac0f207329857a21ba3f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053534.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053535.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053535.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..68cb95a08aea3988e2a15bcfcdf264a67e49633a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053535.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053536.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053536.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d56845454154f7b63b36c3ea37f9313f1116e3f0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053536.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053537.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053537.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f74dd007eb9d757f7af157918281005decfbb376
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053537.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053538.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053538.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d7b47391dcb2d0f2a958ad38d4171382214bf710
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053538.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053539.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053539.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fa8f959aa06be00a32b9bd79556d56c8e8749d20
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053539.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053540.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053540.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2a2e8c507af822e457a5b29f1da7f9a81bb8e9dd
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053540.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053541.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053541.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..63a4ee7eb91f68ade68442547dba65ab43424a5d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053541.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053542.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053542.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6e1fafd1f901ac0c9e5c0740351ec7ee14314140
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053542.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053543.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053543.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5cd573d787976e5a8ebb10568833700b3e470527
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053543.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053544.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053544.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9bb980e956203612574ec2f549e4ea0a36f4e342
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053544.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053545.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053545.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4cf1c4279e08f4d6f20a33d78fe125f7a3532ae5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053545.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053546.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053546.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ca56e0f2d5dde6f6511d117d0db0db8e4207e5a4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053546.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053547.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053547.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..67deed3de3349be60a02d5890a1c774cc8181ca4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053547.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053548.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053548.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..212f7f17c21730b2e0677500270babc563c07297
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053548.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053549.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053549.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9cac7f9761edb897602b88b5bfe16cb088fbf0c0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053549.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053550.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053550.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9a2d195fdf4307cb6e2a0d293359c533cfb235ec
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053550.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053551.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8faaf7191193af03d054da9d09d758093091b4ec
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053552.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053552.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..245d50ec171baefa165c53031774b8c2ebd50509
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053552.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053553.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053553.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..74fb449b41c8ed733885f1c81326981572f9070c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053553.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053554.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053554.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fdc3e6f78b4e4911676ec4732caa84fe38d6fa2e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053554.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053555.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053555.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6119414ff961ce62f3897c9f8f5bc8551d33f990
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053555.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053556.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053556.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..805600ec70f56072091634cadf5f155501c6f996
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053556.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053557.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053557.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3c20b49a263b5647d58a6dfd6595bda2317faf6f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053557.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053558.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053558.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..46c32cf3f1d375af91388cc05a0f590113c54d93
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053558.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053559.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053559.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fcf874733ab311f5124da53a656641f8f6ee11b7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053559.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053560.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053560.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5c21d74fec6b55d0ee843c5a6f30bcbdd9f2e000
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053560.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053561.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053561.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7f20b39b56d3bcc2395f397ccf6d7de3dd2c9ccc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053561.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053562.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053562.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cd81a8b2388a59815ad034cdc0f790cd1493e4e6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053562.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053563.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053563.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cc27bc3fa66fa3fe3e207fe3b9945b752043ee60
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053563.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053564.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053564.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ad773c34c1365983f6db31506218661ef9571208
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053564.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053565.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053565.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fa654fc5fb95072609efba765bd77d797fddcf04
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053565.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053566.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053566.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e2c3bef05451fd613d40fd5de5d7f2f3fac76ad5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053566.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053567.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053567.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2c0f475c8ad660a7756eedc4090097318b70b4fe
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053567.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053568.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053568.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dd6db56ee7f7df971c98147afd6226b0c1ee791b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053568.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053569.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053569.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d5070793c41abd98d9a19dcea4b1a980ca9a737d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053569.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053570.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053570.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6b5ce30082a6a738681ae9ffc32d425c844cef08
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053570.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053571.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053571.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..baaa0e3231009654ce88bf2f73cd634557a8c9fc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053571.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053572.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053572.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2d7ffe9ea92d413933660614aeb958ce5aea2649
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053572.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053573.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053573.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b6aed129d9bb13cb79300f6d2368b2ee18f0c70d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053573.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053574.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053574.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8580bd3308592099465965f28c454849590d1bf4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053574.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053575.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053575.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a0bea23af44b4e2e162c30bf5394489e2fac8450
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053575.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053576.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053576.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c72c8bd6ed9ac55aa4fc369d904c801f2bab1fdb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053576.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053577.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053577.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..368aaa46ea972cec14893d0ed331a5979c8fc9a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053577.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053578.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053578.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..20b43d93cc6e2b26db07886d130779df1bdd7bc8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053578.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053579.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053579.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..da9f08ea05b07e3c21f552269501e70d113aa4c3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053579.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053580.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053580.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f71455f44554b8550c858e055fcb1351a3813452
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053580.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053581.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053581.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..93cb9699b730f43ae813c094549c844d503e1bbc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053581.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053582.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053582.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8f6a89c63fd8b6ee1b23cb41b6c9c170feefb821
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053582.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053583.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053583.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..600a9549c64dcb287a09e5cf4cac41302f23f413
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053583.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053584.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053584.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3626c9c33175a246cd6c04072b02b2a9f8ebed70
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053584.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053585.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053585.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..26ced330d62ca7f706cea70598ca004436d66332
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053585.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053586.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053586.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e7a1ce2bb0f721ca5cae5b3f5582791ebff427ee
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053586.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053587.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053587.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cc76311fe4612d58be5c3ed06207240b9731706f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053587.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053588.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053588.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e2b6fda74c5d42c7c2adb7541aca5c1c636fc53c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053588.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053589.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053589.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..36b0397a6832de7e05495455a61b096c941cb955
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053589.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053590.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053590.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9246c59f836d7157cdc74e9e4d0e563f30307d89
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053590.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053591.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053591.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4641a6d28c333e4fb5fa88b1c8c0ddef42e758b7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053591.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053592.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053592.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a051279e52030786793ccb0d2f7c44f2bb380dc5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053592.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053593.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053593.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..50f3d7de1317f6bcd34f7dda2299412559fd0b32
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053593.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053594.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053594.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7fcac497408e345922e2940989085f052ecedc76
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053594.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053595.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053595.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..763e1e58c01db9a5191b64a87b89cbf1eda7788a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053595.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053596.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053596.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fec64844009948341baafbedf01690a5a7a674be
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053596.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053597.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053597.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6679c1b1ac574fcb77e8ce0ab4cf7d02b82b1f3b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053597.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053598.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053598.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3f9bf9608c55daa4eca0df2ad00542ae9ce2efd8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053598.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053599.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053599.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..11acc9d6e750d72456f235d07127c92ae63f2578
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053599.jpg differ
diff --git a/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053600.jpg b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053600.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fac34bbc8dbafaf69f34c3db6fdcd5a2a7bda7da
Binary files /dev/null and b/data/examples/mosquito_escapes/cam2/20200216_075905-selection/cam2_20200216_0759053600.jpg differ
diff --git a/data/examples/mosquito_escape/cam2_20200303_030120/cam2_20200303_030120-2d_points.csv b/data/examples/mosquito_escapes/cam2/20200303_030120/cam2_20200303_030120-2d_points.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam2_20200303_030120/cam2_20200303_030120-2d_points.csv
rename to data/examples/mosquito_escapes/cam2/20200303_030120/cam2_20200303_030120-2d_points.csv
diff --git a/data/examples/mosquito_escape/cam2_20200303_030120/cam2_20200303_030120-obj1-2d_points.csv b/data/examples/mosquito_escapes/cam2/20200303_030120/cam2_20200303_030120-obj1-2d_points.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam2_20200303_030120/cam2_20200303_030120-obj1-2d_points.csv
rename to data/examples/mosquito_escapes/cam2/20200303_030120/cam2_20200303_030120-obj1-2d_points.csv
diff --git a/data/examples/mosquito_escape/cam2_20200303_030120/cam2_20200303_030120-obj1-3d_track.csv b/data/examples/mosquito_escapes/cam2/20200303_030120/cam2_20200303_030120-obj1-3d_track.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam2_20200303_030120/cam2_20200303_030120-obj1-3d_track.csv
rename to data/examples/mosquito_escapes/cam2/20200303_030120/cam2_20200303_030120-obj1-3d_track.csv
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050001.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0d113628e6d340d4f2aaabe362efe6de5fecdd2c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050051.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1a3bdfb69d5c927f5591455d5270f26a7647fea0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050101.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7d4c313f569c6241db0d75b106b43ec00f486075
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050151.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8451e964fabc653d0397b895e623adee942472db
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050201.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c7f529577eaa6eb8fffb507addb85090ace3ed4c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050251.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9b912549b8953079db0dd71f6848dc49d4d34b23
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050301.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..80360d5127c531368f40cec9cc6c3e31c1149221
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050351.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..fa7d29d9813db6481ef07781dd3ac317fd633f50
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050401.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..64ce03523aa66eacc48bb53e0c7e501214b152d9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050451.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c1b07d802e0a50b236eb9e117451092b81b6308e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050501.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7a353decbdd9f3780412273d57dfcb4feeddaae0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050551.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..de531b63f4e3873e0a75a39372de9df0aabc8a7e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050601.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..33ee7e370141aa87c3b5646cf1acb6d16b14a004
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050651.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7db1508d9c51c090bcbd18a3ead34ca1f8f4f9ea
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050701.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..69366d144fec614d2c3ee5e9ecb059cd19fbe89d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050751.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..45a55fabca21f97711b30a34548d05941c094aa3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050801.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..74f8be46c8b79a702620b0158a4b07cd5a936d82
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050851.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2d87d36fc92625c0d77e49de7009262bb175ee84
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050901.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..52f07dde1dbaad622054880397d7f27c87ae4355
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050951.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d9ae8232d588321ef5de55fcf6ebba85ab4cb137
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759050951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051001.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..87855ebf3d749cfe6b39b8c69ce66d2198a66eec
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051051.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5cf243b81dbfb183cf8f1c55ce738e0fa88f12a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051101.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a66803496cdd9b06d50f3896370430be81124831
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051151.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7ace06023771658fc0682dae6a5d5e4c4ce1bea2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051201.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..55b92da86b77cbf53011a042aebc185ee28aa965
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051251.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1b5bf0dd9858c70e877d885913deb38a3b15aefb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051301.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..de39d85545a077b7040d5a129655b70f4d5731dc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051351.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4116237fba4e6fe69d1c46fe7cd274e7d02b36a5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051401.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..51105ed3214c45eac9f8c69ef88516be84969e65
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051451.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..44330e6bf821dde4aa2eeb46cf4bb7199b06bd69
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051501.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7ccb95635c5e5969be356a30b5c8e0a77a4eced2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051551.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c44bba56f7baf252e1ec5240f2aaba90f9bce8da
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051601.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..74442ea8caf2ee48bb408a33c4314d5db75b9c24
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051651.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..39803d5b47d4667ecbedf47178d28f54aeeca031
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051701.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ad6b2c274abee647361586db090a391381bf5877
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051751.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3743003b083ad781d9ba2eb5d63736c801901b82
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051801.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..87d485fd1f096dd4cff7655b66b88c28bfd29226
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051851.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7e5daedcf896cac03511f125465b676f8062216a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051901.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0ad6ebea856a3dea978b51790e845cba78a6e035
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051951.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..29f91977b574a21b6e100a85ce84cb21a2dc81fa
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759051951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052001.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..707a6100d443105edd20bc0417540e43beaa72f8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052051.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..38594a9df693f389c3546dbf3cace9f14b693f79
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052101.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cdb7f69fb97fcb7e376383de249cb5602c1e976b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052151.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9e1fe0c0839f41b8c4f0c158438fd6c15c4cbfe4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052201.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b5fd1aa435eb810866a347afe0f822c597907b39
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052251.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0da5ddbe20177e77794887368de547cadae0b062
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052301.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cb5aa933485146a1dce303d1ac503e4f81d91783
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052351.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..941c838c81eb27e9dd0867dea2a403d49a9fd8ea
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052401.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ee2cf733b2ce02f70fd9cda45973b63dec98b3ad
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052451.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..336682cc6faf91f92904ff058dca655064fb9f70
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052501.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cf93d81a4e7893ce5c99eb03619a86e71ba58559
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052551.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..315176fe4fdefe36b675918dec51ca6ff5afd429
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052601.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5f083d7f538eaecf531a275f865b11cbe08f388e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052651.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4be4cd53d4ddea2d7b24d6cc6407af6d9fa0f235
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052701.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b783635a03ce677e52bca3bc08203d168ade06ef
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052751.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..37de30692c7e9e10b52d1b05d7485f4c9283757b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052801.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4f7956f6e0b9c8e088538396e180cca099ae57b6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052851.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5e18d458c71da3f07b4bcae6172297c869cf0b99
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052901.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..243a7633c587f761dbe21153ed22fbefa7d3d345
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052951.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7ba865a9896895ac541488155aea4b961259bfc8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759052951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053001.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c6ec9576ed72bc46622f40ae76ebec97a6481687
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053051.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..55b8d665c14b84eb44953fec41dda1cde49633a2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053101.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7ffa50b62fb0017b6776c49835cbfcccb502b40e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053151.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..169919e81ed2287dab936d30347d9533dd630a26
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053201.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7d9a8a9b8ccbe283238286cefe615b27e2935877
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053251.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..84d41bb6391af34f4477b1a505a9ec7fc6b9cf8c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053301.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..36550d26ebb9022b70801b21b2102d5eb655ad7d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053351.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..21334cbdf9c994e5ca3687f4c101dd1322a2e3ba
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053401.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..94e9cfb9dfe04b376631934d37b9c92e2e35ad45
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053451.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ce17b216560ef35ab54740824ba3835e84b57458
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053501.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..81aa969cdfa609c413dacc80c2c2bc285e676db3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053551.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..78ab45cda72c92de72bb3fe6e135db3033e3aeb8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053601.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dc94a583ee89a530b12df824d7276ec33e210d2a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053651.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9787a51aa1074696d23c150ddb318c471161519a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053701.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..477413eb052e273fe800dd1a9425b8140e2b30f1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053751.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f47b67f397e353bc6fd2156fba2537c8850441fc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053801.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c746acf76a96d1a1bda1d1aee937ebf84f8fe03d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053851.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..aa2f4e08e8da115a944a0a65f6c2cc16c61b7fa7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053901.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..725f639502304bfbf9524389d35248da1882c21a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053951.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..178b382cabb76df730451e9858da28ef9739e409
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759053951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054001.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f52d29c1f3df9775d073672676a229216aab3b57
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054051.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3604ff003b1a954f987dc68a42efb44bb7d0f569
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054101.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..88a5db812ea3fcaa5f058efbdc3d892fd6d827ab
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054151.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..126b9dc096950e0cc729e7d25cf35ecd91906dbf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054201.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..011f010d16ab3e7a740cdee58fcb8fe6f84674f5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054251.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8062e64894b815394529bb7a94413bf96331cb5a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054301.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e5b22235f2bc7c7d8c5c5db0ee239adc0fc2af6d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054351.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..30bd8215a3203e45cc24dfc6ec36a0849b6da4ec
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054401.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b385fffcd899c2aa9ec58a06898161bf129cf722
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054451.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4e24fcbf2e8c1a27150eaf2739e7456752060659
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054501.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8db25b50c8d8021a22deb5994ddad3b8014b6149
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054551.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d00e3ecf845bbe86fb2edec993134e51ec17554d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054601.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054601.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..04f164eb71f332be35fee0ed7891c729b6f50215
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054601.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054651.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054651.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..048237e05bdc4fbd54a3e0b8e6a3677b40414ef9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054651.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054701.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054701.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4657e490f09f52db021006ebcf45230d3d14eb51
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054701.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054751.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054751.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5602797f1155a309db8e6e5910a5f44819ef14b8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054751.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054801.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054801.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..2a355aa7a9230755c07d9291198cca5da6875f45
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054801.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054851.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054851.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..27728bed943bd3913f6c0490c060f710a582c051
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054851.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054901.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054901.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7a6af4911a84e9a195438433b058c1a56cc1b3b9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054901.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054951.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054951.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bc5ee8302e7297d5d6bd8bd0fa10ecbe3ac2468c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759054951.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055001.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055001.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9dbd2aa95eaa98b87179a0a3605ea302f0d42ad7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055001.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055051.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055051.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..58d84f9e1f72ebea6cec198666e226e1527156e0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055051.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055101.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055101.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0908cc816c39bcba0d14b4fffaee1d98dd1ddfa2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055101.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055151.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055151.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..bb59131578ca6d42f5054bd3828210591830bf46
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055151.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055201.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055201.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e8e1c3393eaf27fb857d49c5346f1f8bb7a552e6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055201.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055251.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055251.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..812384dd7e6e79d22c17e005383bb51ebdb4340c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055251.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055301.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055301.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f0f9997cf95bdb332c27728041b4147d650164d2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055301.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055351.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055351.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..674384972f740c5b35a965d17b37a8ee62f58b0a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055351.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055401.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055401.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9f28821ef34ff0834f5384fb13eb3ad8ae1f2c68
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055401.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055451.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055451.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b24029d70f414ea39512ec0691f25ad67ec529a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-sample_binned/cam3_20200216_0759055451.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053501.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053501.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..627581df02b44b1aae77d43fd2426a2f8b3d2a18
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053501.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053502.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053502.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5edc13742382a0c4588ca7731a95e11c41410e9e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053502.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053503.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053503.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1a783dd89333cc08c0b8d31aa8e8f5779bbb6ef9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053503.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053504.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053504.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d5f1e1ca4d74297e34312b98bcc02b950b2e0eaa
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053504.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053505.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053505.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..22b62e67fd77e56f7a2aaec374298ddc449f4d42
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053505.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053506.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053506.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3d0832a495cf242aa8ccb2414998bf6b91e6d69e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053506.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053507.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053507.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9b925cd6f96371e218075637a670f771b205d90e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053507.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053508.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053508.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dda70b282dae54032a6581f78e9ae046547e8730
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053508.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053509.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053509.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..adb371fd7f25648b9dab08542602637e2f19601a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053509.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053510.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053510.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1863d5629607baa4c65dfa14c474288a1e89694f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053510.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053511.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053511.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4fa92f2210e45684c293968e15acb33c146a232e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053511.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053512.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053512.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..701d17714541960da1264d5bb095cd31676a5d46
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053512.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053513.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053513.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ae705e6bae3e67ccfd35aff2b0fd3e53c0c2d718
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053513.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053514.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053514.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7c18e5c425c97356a639b10299fa63b7aaec18d3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053514.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053515.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053515.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..86538e388b870a13e2321c7df90d2b6f67ace566
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053515.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053516.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053516.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..394c9bc3d4585d682bed0a3cf24784d338b31902
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053516.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053517.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053517.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d3f9d655751d18109958f49c25bdca9b4eb1f269
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053517.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053518.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053518.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c701409053e6b3969f264020bd1de2575fbd4f51
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053518.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053519.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053519.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..34b376a45b026b5d4eda6619892677f2680934e8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053519.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053520.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053520.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8d64104e6e934864da371dde68d7ca0409849f5a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053520.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053521.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053521.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9b5a1230791084219cc63edab23da5bca52706d7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053521.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053522.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053522.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..88bedf169aea8ce02a1900bd8f342423d9a87db9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053522.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053523.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053523.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d877ef879aa4a79896e8f3c05e468a28c8d5646e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053523.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053524.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053524.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..7b91f0435bdfe41288df253920ee4e5727f02c0c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053524.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053525.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053525.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..18b6268586728b73c83a90cc0ef63d3e31faff3b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053525.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053526.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053526.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8549e87dd3aedc5576752bd43a58706dd8f98514
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053526.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053527.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053527.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c09e315eeeb8bcdc6aeae1d42ccee581e28482df
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053527.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053528.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053528.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..313881b35e2bca63a4256900f5652cbdaa7d38f4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053528.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053529.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053529.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e0b3225073c1b898733cb9e64d3bdc43a042c54c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053529.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053530.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053530.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..636d9379f1efec6a97bf72f07a7d4c7cc315e5a7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053530.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053531.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053531.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..70937f50cb4f4524151f4089add2a78a4c82e4f3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053531.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053532.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053532.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b56a82314a5f73fa202c39bb8e623882227d885b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053532.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053533.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053533.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b4114072f5613d3554f6d735c2fd3d9609c769f4
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053533.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053534.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053534.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..985dd92a69dc523010e590764c04361dca3688a8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053534.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053535.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053535.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0f05bf45c9910a1fd57f5ff992034b548612797e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053535.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053536.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053536.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ab9bfcc041f6327791cb900b92f60268c1a6b783
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053536.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053537.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053537.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..92cca89ebbdac9cd9d7ddab97ab3e1b6c5d78831
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053537.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053538.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053538.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..851395a8a634e0cd2ec427a8c822e182ccf1b2d8
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053538.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053539.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053539.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0d387a8aeafd5f73ae468c8893f8670ccfed1ce3
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053539.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053540.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053540.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..65ffd737b25917790f1b38efb375c7a4deda3c3e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053540.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053541.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053541.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..606f4ede5ef35888064cfecfb26d6d94132d3665
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053541.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053542.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053542.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b2c89a8748fa86bb3fc79b063881d3de78e5c84b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053542.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053543.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053543.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c1ef448f03dc2db58e80249a5c6c2efc55923cb1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053543.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053544.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053544.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c31cdbd6f1069eb8b2046e577212ef5750f333bf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053544.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053545.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053545.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ca848047127906d55ae2f0c2237d32ec402e1886
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053545.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053546.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053546.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..78807448de0d0809c2d4122ab1f3886f63f7910c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053546.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053547.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053547.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b9728a1375efdc2a6869dd0cbfd7b39358bd5c84
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053547.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053548.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053548.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..af8b291a0c86986729bba0af98344c5133186ee0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053548.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053549.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053549.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..8fe7d7f950f723fb63e9c16c3b6c95d8116991f2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053549.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053550.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053550.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..5558827f2bdff8b74b771beb50bf4c1d9349adf7
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053550.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053551.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053551.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dae2609c5d4704248980cdab381fb853816260c6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053551.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053552.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053552.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..be318ccf9c418decf92f984744b5dafa7d683449
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053552.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053553.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053553.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..f106a1b0899f495420ffbb08109263f65a5f7701
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053553.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053554.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053554.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..73b96373dd455924503c8e3c3921f608286be81b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053554.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053555.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053555.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..08e63dc674114b7df8d6df41c4eb44523a8c9717
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053555.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053556.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053556.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4936b7132710bde6b3296680ecc81a7b283ad85b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053556.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053557.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053557.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e02b242f44ac99894605d03a3e7a8c400181d670
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053557.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053558.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053558.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b3d8f9a120c969bdcd1abc63d11683c1a7b98016
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053558.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053559.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053559.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..43ce72a5c2efd7d374437f3f35be9d068d2a3694
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053559.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053560.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053560.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6c224c8f99a16b543b93cfdfda1d033185837218
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053560.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053561.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053561.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..02713887f784affc7e891ca44dcfc83088e487dd
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053561.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053562.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053562.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cf2da61bf5b70069370ea3e2c2bb6222f81a3f7d
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053562.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053563.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053563.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..d15c14d574f2d334164920fe591326ee2b0bd452
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053563.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053564.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053564.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cc2051c76cc69d257890c5955f09a4bc06cbe4ab
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053564.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053565.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053565.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..320536fd159f317674138e193c77f5371330f6e9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053565.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053566.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053566.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..04baacd157486967a8759b89605c2403641f3035
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053566.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053567.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053567.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..63ae3954aae93ad1e144a67bd6a9101a3ee8afcf
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053567.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053568.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053568.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c137c4ed6b27ee0b3f15108720216476ac0f7f76
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053568.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053569.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053569.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..38948cf72d35ed23c8c6be46669f6bd40d3d6eea
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053569.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053570.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053570.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..39bc60a8d0e2bd7b1d123caa100c5e2e02f9d190
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053570.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053571.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053571.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..72a16d4846777f5d377dfb3ff860353faa42e192
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053571.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053572.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053572.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e5a938977d74465decb56b0d17a8309e3933f1b6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053572.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053573.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053573.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..68bd6fc800535eb18b22d5a2e1cc0e948d3c24a9
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053573.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053574.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053574.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..be53bd25cd8fed5af236992ff33230a0802ad3aa
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053574.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053575.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053575.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..495d87d3bbd96ee9fa0db421db07027f5f57e651
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053575.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053576.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053576.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c0399c5c9644830d720186bc300300c510251277
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053576.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053577.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053577.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..4c6bd22e763bf7d71a033631046f97fc5d84936e
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053577.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053578.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053578.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c23f24cc9e453d74947f20e504452d4e22058135
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053578.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053579.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053579.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..549892d90c3d8c68b87c8edbfe72033ff9649922
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053579.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053580.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053580.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e3409feabd467c8205b87e128358292bbaf3d8eb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053580.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053581.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053581.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..9db8bb150b63b2129bcabd6508b517f38326081c
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053581.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053582.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053582.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ffcfc4950c7dbd724f0f046e554a5c1acd7ecc44
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053582.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053583.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053583.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dd62645b21bffcfa012c582ed106175ce3f3b2ae
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053583.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053584.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053584.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ca3b652bedb92e9a73a0eba594584eadbf4763dc
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053584.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053585.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053585.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6b33d28f33240a074e32494458304bebaba45efb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053585.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053586.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053586.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ac5ff64617a711b1b0cd56ebe70eb99f19ea9fc6
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053586.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053587.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053587.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..234525577615420b0f6825a1c532bc7c9908662f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053587.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053588.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053588.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..dd541332ce01341dea79d3e972d160fabc4a45f1
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053588.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053589.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053589.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..077c18e6530418c46f625972e47b834fc2de4047
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053589.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053590.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053590.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..ad7cdfda2e900dc0c2c5ab767ba9e8b2c532ae26
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053590.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053591.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053591.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..b9a0079fb666bd84fb95f1dd4b3ccbdc13107aa5
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053591.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053592.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053592.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..3289b02f5e4f6e7a2a59df79ac776078b509738b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053592.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053593.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053593.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..284411c33c9ded31d1428130cd6194e389d80ec0
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053593.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053594.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053594.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..6f763580ddd70ac79222963e9460e7bb54dadb55
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053594.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053595.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053595.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..44e6016e0ba2b1be6faaa633cfcecd514ea6afeb
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053595.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053596.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053596.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..06fe97fd9c71aa3457ac01d4889eb48a62efa04b
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053596.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053597.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053597.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..a16216dd1bb4246ac6e7a81e476b9f0e1b6f4108
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053597.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053598.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053598.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..441f6fa7acdfea286c1063af1aaa714a58faf03f
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053598.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053599.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053599.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..1a24c6246ff78692a765267c901844d97ea02d7a
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053599.jpg differ
diff --git a/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053600.jpg b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053600.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..0afdab340300b164d639b9c6a295dfe8adf8ccc2
Binary files /dev/null and b/data/examples/mosquito_escapes/cam3/20200216_075905-selection/cam3_20200216_0759053600.jpg differ
diff --git a/data/examples/mosquito_escape/cam3_20200303_030120/cam3_20200303_030120-2d_points.csv b/data/examples/mosquito_escapes/cam3/20200303_030120/cam3_20200303_030120-2d_points.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam3_20200303_030120/cam3_20200303_030120-2d_points.csv
rename to data/examples/mosquito_escapes/cam3/20200303_030120/cam3_20200303_030120-2d_points.csv
diff --git a/data/examples/mosquito_escape/cam3_20200303_030120/cam3_20200303_030120-obj1-2d_points.csv b/data/examples/mosquito_escapes/cam3/20200303_030120/cam3_20200303_030120-obj1-2d_points.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam3_20200303_030120/cam3_20200303_030120-obj1-2d_points.csv
rename to data/examples/mosquito_escapes/cam3/20200303_030120/cam3_20200303_030120-obj1-2d_points.csv
diff --git a/data/examples/mosquito_escape/cam3_20200303_030120/cam3_20200303_030120-obj1-3d_track.csv b/data/examples/mosquito_escapes/cam3/20200303_030120/cam3_20200303_030120-obj1-3d_track.csv
similarity index 100%
rename from data/examples/mosquito_escape/cam3_20200303_030120/cam3_20200303_030120-obj1-3d_track.csv
rename to data/examples/mosquito_escapes/cam3/20200303_030120/cam3_20200303_030120-obj1-3d_track.csv
diff --git a/images/__init__.py b/images/__init__.py
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..b74215e0fbbb6512474bdc2ed9cd9b75eff4ad43 100644
--- a/images/__init__.py
+++ b/images/__init__.py
@@ -0,0 +1,3 @@
+from images.image import Image, TypeImage
+from images.imagesequence import ImageSequence
+from images.multiimagesequence import MultiImageSequence
\ No newline at end of file
diff --git a/images/image.py b/images/image.py
new file mode 100644
index 0000000000000000000000000000000000000000..00d38089f4805a16987a2947f63b94d9fb97abc4
--- /dev/null
+++ b/images/image.py
@@ -0,0 +1,622 @@
+"""Images."""
+
+from abc import ABCMeta, abstractmethod
+from enum import Enum
+
+import os
+from typing import Any, Optional, Tuple, Union
+
+import numpy as np
+import PIL.Image
+
+__author__ = "C.J. Voesenek"
+__maintainer__ = "C.J. Voesenek"
+__email__ = "cees.voesenek@wur.nl"
+
+
+class TypeImage(Enum):
+ """Image types.
+
+ Options:
+ BW: Black and white (i.e. binary).
+ GREY: Greyscale images.
+ RGB: RGB images.
+ """
+ BW = "bw"
+ GREY = "grey"
+ RGB = "rgb"
+
+
+class TypeFileImage(Enum):
+ """Image file types.
+
+ Options:
+ TIFF_BW: Black and white TIFFs.
+ TIFF8: 8-bit TIFFs.
+ TIFF16: 16-bit TIFFs.
+ """
+ TIFF_BW = "tiff_bw"
+ TIFF8 = "tiff8"
+ TIFF16 = "tiff16"
+
+
+class OperatorLogical(Enum):
+ """Logical operator types.
+
+ Options:
+ NOT
+ AND
+ OR
+ """
+ NOT = "not"
+ AND = "and"
+ OR = "or"
+
+
+class OperatorComparison(Enum):
+ """Comparison operator types.
+
+ Options:
+ EQUALS
+ LESS
+ GREATER
+ LESS_EQUALS
+ GREATER_EQUALS
+ """
+ EQUALS = "="
+ LESS = "<"
+ GREATER = ">"
+ LESS_EQUALS = "<="
+ GREATER_EQUALS = ">="
+
+
+class _ImageData(metaclass=ABCMeta):
+ """Abstract image data."""
+ @property
+ @abstractmethod
+ def nb_xi(self) -> int: pass
+
+ @property
+ @abstractmethod
+ def nb_eta(self) -> int: pass
+
+ @property
+ @abstractmethod
+ def shape(self) -> int: pass
+
+ @property
+ @abstractmethod
+ def type(self) -> TypeImage: pass
+
+ @property
+ @abstractmethod
+ def get(self) -> np.ndarray: pass
+
+ @property
+ @abstractmethod
+ def clear_cache(self) -> None: pass
+
+ @property
+ @abstractmethod
+ def use_caching(self) -> bool: pass
+
+ @use_caching.setter
+ @abstractmethod
+ def use_caching(self, use_caching: bool) -> None: pass
+
+
+class _ImageDataArray(_ImageData, metaclass=ABCMeta):
+ """Images data as an array."""
+
+ def __init__(self, data: Optional[np.ndarray] = None,
+ shape: Optional[Tuple[int, int]] = None,
+ type_data: TypeImage = TypeImage.GREY) -> None:
+ """Initialises image data from scratch."""
+ self._type = type_data # type: TypeImage
+
+ self._data = None # type: np.ndarray
+ if data is not None and shape is None:
+ self._data = data
+ elif shape is not None and data is None:
+ self._initialise_array(shape)
+ else:
+ raise ValueError("Specify either data or shape.")
+
+ @property
+ def nb_xi(self) -> int:
+ return self._data.shape[0]
+
+ @property
+ def nb_eta(self) -> int:
+ return self._data.shape[1]
+
+ @property
+ def shape(self) -> Tuple[int, int]:
+ return self._data.shape
+
+ @property
+ def type(self) -> TypeImage:
+ return self._type
+
+ @property
+ def use_caching(self) -> bool:
+ raise RuntimeError("Array-based images do not have a cache.")
+
+ @use_caching.setter
+ def use_caching(self, use_caching: bool) -> None:
+ raise RuntimeError("Array-based images do not have a cache.")
+
+ def clear_cache(self) -> None:
+ raise RuntimeError("Array-based images do not have a cache, "
+ "so it cannot be cleared.")
+
+ def get(self) -> np.ndarray:
+ return self._data
+
+ @abstractmethod
+ def _initialise_array(self, shape: Tuple[int, int]): pass
+
+ @staticmethod
+ def build(data: np.ndarray = None, shape: Tuple[int, int] = None,
+ type_data: TypeImage = TypeImage.GREY) -> "_ImageDataArray":
+ if type_data == TypeImage.GREY:
+ return _ImageDataGrey(data, shape)
+ elif type_data == TypeImage.BW:
+ return _ImageDataBW(data, shape)
+ elif type_data == TypeImage.RGB:
+ raise NotImplementedError("RGB images have not yet been "
+ "implemented.")
+
+
+class _ImageDataGrey(_ImageDataArray):
+ """Greyscale image data."""
+ def __init__(self, data: Optional[np.ndarray] = None,
+ shape: Optional[Tuple[int, int]] = None) -> None:
+ """Initialises greyscale image data."""
+ super(_ImageDataGrey, self).__init__(data, shape,
+ type_data=TypeImage.GREY)
+
+ def _initialise_array(self, shape: Tuple[int, int]) -> None:
+ self._data = np.zeros(shape)
+
+
+class _ImageDataBW(_ImageDataArray):
+ """Black and white image data."""
+ def __init__(self, data: Optional[np.ndarray] = None,
+ shape: Optional[Tuple[int, int]] = None) -> None:
+ """Initialises greyscale image data."""
+ super(_ImageDataBW, self).__init__(data, shape,
+ type_data=TypeImage.BW)
+
+ def _initialise_array(self, shape: Tuple[int, int]) -> None:
+ self._data = np.zeros(shape, dtype=np.bool_)
+
+
+class _ImageDataFile(_ImageData):
+ """Image data from a file."""
+
+ def __init__(self, file: str, type_data: TypeImage,
+ use_caching: bool = True) -> None:
+ """Initialises image data from a file."""
+ self._file = file # type:
+ self._use_caching = use_caching # type: bool
+ self._cache = None # type: np.ndarray
+ self._type = type_data # type: TypeImage
+ self._shape = None # type: Tuple[int, int]
+
+ self._check_existence()
+
+ @property
+ def nb_xi(self) -> int:
+ if self._shape is None:
+ self._load_shape()
+ return self._shape[0]
+
+ @property
+ def nb_eta(self) -> int:
+ if self._shape is None:
+ self._load_shape()
+ return self._shape[1]
+
+ @property
+ def shape(self) -> Tuple[int, int]:
+ if self._shape is None:
+ self._load_shape()
+ return self._shape
+
+ @property
+ def type(self) -> TypeImage:
+ return self._type
+
+ @property
+ def use_caching(self) -> bool:
+ return self._use_caching
+
+ @use_caching.setter
+ def use_caching(self, use_caching: bool) -> None:
+ self._use_caching = use_caching
+
+ def get(self, cache: bool = True) -> np.ndarray:
+ """Gets the image data.
+
+ Args:
+ cache: Whether to save the image data in the cache.
+
+ Returns:
+ The image data.
+ """
+ if self._cache is None:
+ data = self._load_image()
+ return data
+ else:
+ return self._cache
+
+ def clear_cache(self) -> None:
+ """Clears the cached image."""
+ self._cache = None
+
+ def _load_image(self) -> np.ndarray:
+ """Loads the image from the specified file.
+
+ Depending on whether caching is used, this also caches the file.
+
+ Returns:
+ The image data.
+ """
+ if self._type == TypeImage.GREY:
+ image = np.array(PIL.Image.open(self._file))
+ if image.dtype == np.uint8:
+ data = image / (2 ** 8 - 1.0)
+ elif image.dtype == np.uint16:
+ data = image / (2 ** 16 - 1.0)
+ else:
+ raise ValueError("Unknown datatype in file \"{}\" for "
+ "greyscale images, can only convert 8-bit "
+ "and 16-bit images.".format(self._file))
+ self._shape = data.shape
+ elif self._type == TypeImage.BW:
+ data = np.array(PIL.Image.open(self._file))
+ if data.dtype != "bool":
+ raise ValueError("The specified image \"{}\" is not black and "
+ "white.".format(self._file))
+ else:
+ raise NotImplementedError("Other data types than grayscale for "
+ "loading images are not yet implemented.")
+
+ if self._use_caching:
+ self._cache = data
+ return data
+
+ def _load_shape(self) -> None:
+ """Reads the dimensions from the image."""
+ image = np.array(PIL.Image.open(self._file))
+ self._shape = image.shape
+
+ def _check_existence(self) -> None:
+ """Checks whether the specified file exists.
+
+ Raises:
+ IOError: File {} does not exist.
+ """
+ if not os.path.isfile(self._file):
+ raise IOError("File {} does not exist.".format(self._file))
+
+
+class Image:
+ """An image.
+
+ Attributes:
+ nb_xi: The number of points along the first image dimension.
+ nb_eta: The number of points along the second image dimension.
+ shape: The number of points along respectively the xi- and eta-
+ dimensions.
+ """
+
+ def __init__(self, **kwargs) -> None:
+ """Initialises an image.
+
+ Images can be loaded from a file, in which case they are optionally
+ cached, allowing you to free memory when the image is not in use
+ anymore. Images can also be created from scratch, in which case they
+ will remain in memory.
+
+ Args:
+ file: The path to an image file to load. Cannot be used in
+ combination with "shape".
+ image_type: The type of image to load.
+ use_caching: Whether to use caching or load the image
+ immediately. Only useable when "file" has been specified.
+ shape: The shape of the image to create. Cannot be used in
+ combination with "file".
+ data: The array containing the image.
+
+ Raises:
+ ValueError: Invalid keyword argument, for file images, use only
+ "type" and "use_cache".
+ ValueError: Invalid keyword argument, for array images, use only
+ "type".
+ """
+ self._data = None # type: _ImageData
+
+ if "file" in kwargs:
+ # Check for invalid keyword arguments
+ for arg in kwargs:
+ if arg not in ("file", "type", "use_caching"):
+ raise ValueError("Invalid keyword argument, for file "
+ "images, use only \"type\" and "
+ "\"use_cache\".")
+ file = kwargs["file"]
+ if "type" not in kwargs:
+ type_image = TypeImage.GREY
+ else:
+ type_image = TypeImage(kwargs["type"])
+ if "use_caching" not in kwargs:
+ use_caching = True
+ else:
+ use_caching = kwargs["use_caching"]
+
+ self._data = _ImageDataFile(file, type_image, use_caching)
+ elif "shape" in kwargs:
+ for arg in kwargs:
+ if arg not in ("shape", "type"):
+ raise ValueError("Invalid keyword argument, for array "
+ "images, use only \"shape\" and \"type\".")
+ shape = kwargs["shape"]
+ if "type" not in kwargs:
+ type_image = TypeImage.GREY
+ else:
+ type_image = TypeImage(kwargs["type"])
+ self._data = _ImageDataArray.build(shape=shape,
+ type_data=type_image)
+ elif "data" in kwargs:
+ for arg in kwargs:
+ if arg not in ("data", "type"):
+ raise ValueError("Invalid keyword argument, "
+ "for directly specified array "
+ "images, use only \"data\" and \"type\".")
+ data = kwargs["data"]
+ if "type" not in kwargs:
+ type_image = TypeImage.GREY
+ else:
+ type_image = TypeImage(kwargs["type"])
+ self._data = _ImageDataArray.build(data=data,
+ type_data=type_image)
+ else:
+ raise ValueError("Specify either \"file\" or \"shape\" as "
+ "keywords, for images from file or from scratch "
+ "respectively.")
+
+ @property
+ def nb_xi(self) -> int:
+ return self._data.nb_xi
+
+ @property
+ def nb_eta(self) -> int:
+ return self._data.nb_eta
+
+ @property
+ def shape(self) -> Tuple[int, int]:
+ return self._data.shape
+
+ @property
+ def type(self) -> TypeImage:
+ return self._data.type
+
+ @property
+ def use_caching(self) -> bool:
+ return self._data.use_caching
+
+ @use_caching.setter
+ def use_caching(self, use_caching) -> None:
+ self._data.use_caching = use_caching
+
+ def clear_cache(self) -> None:
+ """Clears the cached data.
+
+ This function will raise an error if caching is not available.
+ """
+ self._data.clear_cache()
+
+ def copy(self) -> "Image":
+ """Copies the image.
+
+ Note that this will turn file-based images into array-based images;
+ the data will be loaded.
+
+ Returns:
+ A physical copy of the image - the array is duplicated in memory.
+ """
+ return Image(data=np.copy(self.get()), type=self.type)
+
+ def get(self) -> np.ndarray:
+ """Gets the image data as an array.
+
+ Returns:
+ An array with the image data.
+ """
+ return self._data.get()
+
+ def gradient(self, return_components: bool = False) \
+ -> Union["Image", Tuple["Image", "Image"]]:
+ """Calculates the image's spatial gradient.
+
+ Args:
+ return_components: Whether to return the components (if True) or a
+ magnitude (if False).
+ """
+ if self.type != TypeImage.GREY:
+ raise ValueError("Can only calculate spatial gradients for "
+ "greyscale images.")
+
+ grad = np.gradient(self.get(), edge_order=2)
+ if return_components:
+ return Image(data=grad[0], type="grey"), \
+ Image(data=grad[1], type="grey")
+ else:
+ return Image(data=np.sqrt(grad[0]**2 + grad[1]**2),
+ type="grey")
+
+ def logical(self, operation: str,
+ other: Optional["Image"] = None) -> "Image":
+ """Performs a logical operation on the image.
+
+ Note that this can only be used on black and white images.
+
+ Returns:
+ The image with the operation performed on it.
+ """
+ if self.type != TypeImage.BW or \
+ (other is not None and other.type != TypeImage.BW):
+ raise ValueError("Logical operations can only be performed on "
+ "black and white images.")
+
+ operation = OperatorLogical(operation)
+
+ if operation == OperatorLogical.NOT:
+ return Image(data=np.logical_not(self.get()), type=TypeImage.BW)
+ elif operation == OperatorLogical.AND:
+ if other is None:
+ raise ValueError("Specify another operand to perform the "
+ "logical AND with.")
+ return Image(data=np.logical_and(self.get(), other.get()),
+ type=TypeImage.BW)
+ elif operation == OperatorLogical.OR:
+ if other is None:
+ raise ValueError("Specify another operand to perform the "
+ "logical OR with.")
+ return Image(data=np.logical_or(self.get(), other.get()),
+ type=TypeImage.BW)
+
+ def normalise(self, minimum: Optional[float] = None,
+ maximum: Optional[float] = None):
+ """Normalises the image to the interval [0.0, 1.0].
+
+ Note: values below the minimum and above the maximum will be clipped
+ to 0.0 and 1.0 respectively.
+
+ Args:
+ minimum: The minimum value, i.e. the value that will represent 0.0.
+ If None is specified, it will use the minimum value in the
+ image.
+ maximum: The maximum value, i.e. the value that will represent 1.0.
+ If None is specified, it will use the maximum value in the
+ image.
+ """
+ if self.type != TypeImage.GREY:
+ raise ValueError("Can only normalise greyscale images.")
+
+ if minimum is None:
+ minimum = np.min(self.get())
+ if maximum is None:
+ maximum = np.max(self.get())
+ data = (self.get() - minimum) / (maximum - minimum)
+
+ # Clip values outside [0.0, 1.0].
+ data[data < 0.0] = 0.0
+ data[data > 1.0] = 1.0
+
+ return Image(data=data, type=self.type)
+
+ def threshold(self, threshold: float, operator_comparison: str = ">") \
+ -> "Image":
+ """Thresholds a greyscale image."""
+ if self.type != TypeImage.GREY:
+ raise ValueError("Can only threshold greyscale images.")
+
+ operator_comparison = OperatorComparison(operator_comparison)
+ if operator_comparison == OperatorComparison.EQUALS:
+ return Image(data=self.get() == threshold, type="bw")
+ elif operator_comparison == OperatorComparison.LESS:
+ return Image(data=self.get() < threshold, type="bw")
+ elif operator_comparison == OperatorComparison.GREATER:
+ return Image(data=self.get() > threshold, type="bw")
+ elif operator_comparison == OperatorComparison.LESS_EQUALS:
+ return Image(data=self.get() <= threshold, type="bw")
+ elif operator_comparison == OperatorComparison.GREATER_EQUALS:
+ return Image(data=self.get() >= threshold, type="bw")
+
+ def write(self, file: str, type_file: str = "tiff16") -> None:
+ """Writes the image to a file.
+
+ Args:
+ file: The file to write to.
+ type_file: The file type to write.
+ """
+ type_file = TypeFileImage(type_file)
+ if self.type == TypeImage.GREY:
+ if np.min(self.get()) < 0.0 or np.max(self.get()) > 1.0:
+ raise ValueError("The current image has values outside the "
+ "interval [0.0, 1.0], and cannot be written.")
+
+ if self.type == TypeImage.GREY:
+ if type_file == TypeFileImage.TIFF8:
+ arr = (self.get() * (2**8 - 1)).astype(dtype=np.uint8)
+ elif type_file == TypeFileImage.TIFF16:
+ arr = (self.get() * (2**16 - 1)).astype(dtype=np.uint16)
+ else:
+ raise ValueError("Can only write greyscale images to \"tiff8\" "
+ "and \"tiff16\" files.")
+ image = PIL.Image.fromarray(arr)
+ image.save(file, format="TIFF")
+ elif self.type == TypeImage.BW:
+ if type_file != TypeFileImage.TIFF_BW:
+ raise ValueError("Can only write B/W images to \"tiff_bw\" ""files.")
+
+ # FIXME: Directly converting from a boolean does not work. We use an
+ # inefficient workaround here. This should be fixed at some point.
+ arr = np.zeros(self.shape, dtype=np.uint8)
+ arr[self.get()] = 255
+ image = PIL.Image.fromarray(arr).convert("1")
+ image.save(file, format="TIFF")
+
+ def __add__(self, other: Union[float, "Image"]) -> "Image":
+ self._check_valid_for_operation("addition", other)
+
+ if type(other) is float or type(other) is int:
+ data = self.get() + other
+ else:
+ data = self.get() + other.get()
+ return Image(data=data, type=TypeImage.GREY)
+
+ def __sub__(self, other: Union[float, int, "Image"]) -> "Image":
+ self._check_valid_for_operation("subtraction", other)
+
+ if type(other) is float or type(other) is int:
+ data = self.get() - other
+ else:
+ data = self.get() - other.get()
+ return Image(data=data, type=TypeImage.GREY)
+
+ def __mul__(self, other: Union[float, int]) -> "Image":
+ self._check_valid_for_operation("multiplication", other)
+
+ data = self.get() * other
+ return Image(data=data, type=TypeImage.GREY)
+
+ def __truediv__(self, other: Union[float, int]) -> "Image":
+ self._check_valid_for_operation("division", other)
+
+ data = self.get() / other
+ return Image(data=data, type=TypeImage.GREY)
+
+ def __eq__(self, other: "Image") -> bool:
+ return np.array_equal(self.get(), other.get())
+
+ def _check_valid_for_operation(self, operation: str, other: Any) -> None:
+ if operation in ("addition", "subtraction"):
+ if not(type(other) is Image or type(other) is float or
+ type(other) is int):
+ raise ValueError("Can only perform {} with other images or "
+ "floating point numbers.".format(operation))
+ if type(other) is Image:
+ if self.type != TypeImage.GREY or other.type != TypeImage.GREY:
+ raise ValueError("Can only perform {} on grayscale "
+ "images.".format(operation))
+ if self.shape != other.shape:
+ raise ValueError("Can only perform {} on images with the "
+ "same shape.".format(operation))
+ elif operation in ("multiplication", "division"):
+ if not (type(other) is float or type(other) is int):
+ raise ValueError("Can only perform {} with a "
+ "float or int (which will be treated "
+ "as a float).".format(operation))
diff --git a/images/imagesequence.py b/images/imagesequence.py
new file mode 100644
index 0000000000000000000000000000000000000000..f11a6c3305ec98fced1e7a7484fd3638eb766313
--- /dev/null
+++ b/images/imagesequence.py
@@ -0,0 +1,278 @@
+import os
+import re
+from typing import List, Tuple
+
+import numpy as np
+
+from images import Image, TypeImage
+from images.utils import find_common_substrings_with_one_separator, find_uncommon_substrings, find_file_numbers
+
+
+__author__ = "C.J. Voesenek and A. Cribellier"
+__maintainer__ = "A. Cribellier"
+__email__ = "antoine.cribellier@wur.nl"
+
+
+class ImageSequence:
+ """Sequence of images on disk.
+
+ This object can be indexed like an array (i.e. sequence[0]), retrieving
+ the nth image from the sequence.
+ """
+
+ def __init__(self, images: List[Image], names: List[str] = None, path: str = None, check: bool = False) -> None:
+ """Initialises an ImageSequence object.
+
+ Args:
+ images: The images that this sequence comprises.
+ names: Names of the images in the sequence
+ path: The path of the directory with the images
+ check: Check the images for consistency within the sequence.
+ """
+
+ if names is None:
+ names = ['img{}'.format(x) for x in range(1, len(images)+1)]
+
+ self._images = images # type: List[Image]
+ self._names = names # type: List[str]
+ self._path = path # type: str
+
+ # Will try to get image numbers from their names
+ common_substrings_dict = find_common_substrings_with_one_separator(self._names)
+ most_common_substring = next(iter(common_substrings_dict)) # get first key
+ num_occurrence = common_substrings_dict[most_common_substring]
+ if num_occurrence == len(self._names):
+ # uncommon_substrings = find_uncommon_substrings(self._names, most_common_substring)
+ # self._numbers = [int(s) for s in uncommon_substrings] # type: List[int]
+ self._numbers = find_file_numbers(self._names, most_common_substring) # type: List[int]
+ assert len(self._numbers) == len(self._names)
+
+ else:
+ self._numbers = list(range(1, len(self._names)+1)) # type: List[int]
+
+ if check:
+ self._check_images()
+
+ @property
+ def nb_images(self) -> int:
+ return len(self._images)
+
+ @property
+ def nb_xi(self) -> int:
+ return self[0].nb_xi
+
+ @property
+ def nb_eta(self) -> int:
+ return self[0].nb_eta
+
+ @property
+ def shape(self) -> Tuple[int, int]:
+ return self[0].shape
+
+ @property
+ def type(self) -> TypeImage:
+ return self[0].type
+
+ def get(self, index: int) -> Image:
+ """Retrieves an image from the sequence.
+
+ Args:
+ index: The index of the image to load.
+
+ Returns:
+ The image at the specified index.
+ """
+ return self._images[index]
+
+ def get_name(self, index: int) -> str:
+ """Retrieves the name of an image from the sequence.
+
+ Args:
+ index: The index of the image.
+
+ Returns:
+ The name of the image at the specified index.
+ """
+ return self._names[index]
+
+ def get_names(self) -> List[str]:
+ """Retrieves the names of all the image of the sequence.
+
+ Returns:
+ The names of the images
+ """
+ return self._names
+
+ def get_numbers(self) -> List[int]:
+ """Retrieves the numbers of all the image of the sequence.
+
+ Returns:
+ The numbers of the images (e.g. frame number)
+ """
+ return self._numbers
+
+ def get_path(self) -> List[str]:
+ """Retrieves the path of the directory where the images of the sequence are.
+
+ Returns:
+ The path of the directory with the images.
+ """
+ return self._path
+
+ def minimum(self, use_caching: bool = None) -> Image:
+ """Calculates the minimum of each pixel in the image sequence.
+
+ Args:
+ use_caching: Whether to use caching while loading all images to
+ calculate the minimum. This will use a lot of memory,
+ but will save loading times if multiple operations are
+ performed.
+
+ Returns:
+ The minimum image of this sequence.
+ """
+ change_caching = use_caching is not None
+
+ image_min = self[0].get()
+ for image in self:
+ if change_caching:
+ use_caching_old = image.use_caching
+ image.use_caching = use_caching
+ image_min = np.minimum(image.get(), image_min)
+ if change_caching:
+ image.use_caching = use_caching_old
+ return Image(data=image_min, type=self.type)
+
+ def maximum(self, use_caching: bool = None) -> Image:
+ """Calculates the maximum of each pixel in the image sequence.
+
+ Args:
+ use_caching: Whether to use caching while loading all images to
+ calculate the maximum. This will use a lot of memory,
+ but will save loading times if multiple operations are
+ performed.
+
+ Returns:
+ The maximum image of this sequence.
+ """
+ change_caching = use_caching is not None
+
+ image_max = self[0].get()
+ for image in self:
+ if change_caching:
+ use_caching_old = image.use_caching
+ image.use_caching = use_caching
+ image_max = np.maximum(image.get(), image_max)
+ if change_caching:
+ image.use_caching = use_caching_old
+ return Image(data=image_max, type=self.type)
+
+ def median(self) -> Image:
+ """Calculates the median of each pixel in the image sequence.
+
+ Returns:
+ The median image of this sequence.
+ """
+ images = np.zeros((self.nb_images, ) + self.shape)
+ for i in range(self.nb_images):
+ # Do not use caching here, as it would use twice the memory for
+ # an already memory-intensive operation.
+ use_caching_old = self[i].use_caching
+ self[i].use_caching = False
+
+ images[i, :, :] = self[i].get()
+
+ self[i].use_caching = use_caching_old
+
+ return Image(data=np.median(images, axis=0), type=self.type)
+
+ def __getitem__(self, index: int) -> Image:
+ return self.get(index)
+
+ def delete(self, index) -> None:
+ """Deletes an image from the sequence.
+ """
+
+ self._images.pop(index)
+ self._names.pop(index)
+
+ def _check_images(self) -> None:
+ shape_ref = self[0].shape
+ type_ref = self[0].type
+ for image in self:
+ if image.shape != shape_ref or image.type != type_ref:
+ raise ValueError("All images in the sequence should have the "
+ "same size and type.")
+
+ @staticmethod
+ def from_directory(directory: str,
+ expr: str = r"\d+(?=\.(tif|png|jpg|bmp))",
+ identifiers: List[int] = None,
+ use_caching: bool = True,
+ check: bool = False,
+ type_image: str = "grey") -> "ImageSequence":
+ """Gathers image files from directory.
+
+ This function sets the files attribute of the instance to a list of
+ files from the specified directory, matching the regular expression
+ and optionally the image identifiers. If no image identifiers are
+ specified, the list is in alphabetical order. If image identifiers
+ are specified, the output is sorted in the same order
+
+ Args:
+ directory: The path to a directory containing images.
+ expr: Regular expression pattern matching image files.
+ identifiers: The identifiers of the images that should be loaded
+ from this sequence.
+ use_caching: Whether or not to use caching for the images.
+ check: Whether or not to check all images for correctness.
+ type_image: The type of the images ('rgb', 'grey' or 'bw').
+
+ Raises:
+ IOError: Could not match images to all specified identifiers.
+ IOError: No image files matching {expr} found in directory
+ {directory}.
+ """
+ contents = os.listdir(directory)
+ pattern = re.compile(expr)
+
+ files = []
+ file_names = []
+ id_list = []
+ for file_name in contents:
+ match = re.search(pattern, file_name)
+ if match:
+ # Check whether this file is in the (optionally specified) list
+ # of identifiers
+ if identifiers:
+ id_cur = int(match.group(0))
+ do_append = id_cur in identifiers
+ if do_append:
+ id_list.append(id_cur)
+ else:
+ do_append = True
+
+ if do_append:
+ files.append(os.path.join(directory, file_name))
+ file_names.append(os.path.basename(os.path.normpath(file_name)))
+
+ # If a list of identifiers was specified, check whether we found
+ # them all, then sort the list by identifier
+ if identifiers is not None:
+ if len(identifiers) != len(files):
+ raise IOError("Could not match images to all specified "
+ "identifiers.")
+ files.sort(key=dict(zip(files, id_list)).get)
+ else:
+ files.sort()
+
+ if len(files) == 0:
+ raise IOError("No image files matching \"" + expr +
+ "\" found in directory \"" + directory + "\".")
+
+ images = []
+ for file in files:
+ images.append(Image(file=file, type=type_image,
+ use_caching=use_caching))
+
+ return ImageSequence(images, file_names, directory, check=check)
diff --git a/images/multiimagesequence.py b/images/multiimagesequence.py
new file mode 100644
index 0000000000000000000000000000000000000000..c44c9d620fa2a93a30c5dae3e31b3f0ffe10617c
--- /dev/null
+++ b/images/multiimagesequence.py
@@ -0,0 +1,278 @@
+import os
+from math import gcd
+from typing import Dict, List, Optional, Tuple
+
+import numpy as np
+
+from images import Image, ImageSequence
+
+__author__ = "C.J. Voesenek and A. Cribellier"
+__maintainer__ = "A. Cribellier"
+__email__ = "antoine.cribellier@wur.nl"
+
+
+class MultiImageSequence:
+ """A combination of multiple synchronised image sequences (a.k.a a recording).
+
+ Attributes:
+ names: List of names of the sequences.
+ sequences: A dictionary of image sequences, with the names as keys.
+ nb_sequences: The number of image sequences in this object.
+
+ index: A dictionary of lists with the indices into an image sequence
+ of each overall time step. When images are not defined at a time
+ instant, it contains None.
+
+ frame_rate_total: The overall frame rate of this multi-camera sequence
+ frame_rates: A dictionary of the frame rates of every sequence,
+ with the identifiers as keys.
+
+ time: A list of times at every step.
+ nb_time: The number of time steps in this sequence.
+ """
+
+ def __init__(self, sequences: Dict[str, ImageSequence],
+ frame_rates: Dict[str, float] or int,
+ shifts: Dict[str, int] = None,
+ paths: List[str] = None) -> None:
+ """Initialises a MultiImageSequence.
+
+ Args:
+ sequences: A dictionary of image sequences, with the identifiers
+ as keys.
+ frame_rates: A dictionary of the frame rates of every sequence,
+ with the identifiers as keys or an int if the frame_rate is the same for all.
+ shifts: A dictionary of shifts in number of frames of the start of
+ each sequence with respect to a common time point, with the
+ identifiers as keys. This is used when cameras are running at a
+ different frame rate, but the video sequences do not start at
+ the exact same time instance (as could happen when e.g.,
+ and end trigger is used). For example, a camera running at
+ half the frame rate of another camera may have a shift of one
+ frame w.r.t. the other.
+ paths: List of paths to directories containing each with a recording sequence
+
+ Raises:
+ ValueError: Specify all image sequences as an ImageSequence object.
+ """
+
+ if type(frame_rates) is int:
+ frame_rates = {identifier: frame_rates for identifier in self.names}
+
+ # Check whether the dictionaries have matching identifiers.
+ valid_keys = True
+ for name in sequences.keys():
+ valid_keys = valid_keys and name in frame_rates
+ if shifts is not None:
+ valid_keys = valid_keys and name in shifts
+
+ if not valid_keys:
+ raise ValueError("The dictionaries of image sequences, "
+ "frame rates, and (optionally) shifts should "
+ "all have the same keys")
+
+ # Check whether all sequences are ImageSequence objects
+ for key in sequences.keys():
+ if not isinstance(sequences[key], ImageSequence):
+ raise ValueError("Specify all image sequences as an "
+ "ImageSequence object.")
+
+ self.sequences = sequences # type: Dict[str, ImageSequence]
+
+ self.frame_rate_total = None # type: float
+ self.frame_rates = frame_rates # type: Dict[str, float]
+ self.index = None # type: Dict[str, List[int]]
+ self.time = None # type: np.ndarray
+
+ if not shifts:
+ shifts = dict()
+ for name in self.names:
+ shifts[name] = 0
+ self.shifts = shifts # type: Dict[str, int]
+
+ if paths is not None:
+ assert len(paths) == len(self.names)
+ self._paths = paths # type: List[str]
+
+ self._calculate_time()
+
+ @property
+ def names(self) -> List[str]:
+ return list(self.sequences.keys())
+
+ @property
+ def nb_sequences(self) -> int:
+ return len(self.sequences)
+
+ @property
+ def nb_time(self) -> int:
+ return self.time.size
+
+ def _calculate_time(self) -> None:
+ """Calculates a time vector and the sequence indices in it."""
+
+ # Find the greatest common divisor of all frame rate combinations. We
+ # progress through the list, updating the overall frame rate every
+ # time using the greatest common divisor (gcd()).
+ self.frame_rate_total = self.frame_rates[self.names[0]]
+ for name in self.names:
+ div = gcd(self.frame_rate_total, self.frame_rates[name])
+ self.frame_rate_total = div * (self.frame_rate_total //
+ div * self.frame_rates[name] // div)
+
+ # Calculate step size with respect to the overall frame rate. Also
+ # adjust the shift to be in the overall time vector.
+ step = dict(zip(self.names, [0] * self.nb_sequences))
+ nb_t = dict(zip(self.names, [0] * self.nb_sequences))
+ for name in self.names:
+ step[name] = self.frame_rate_total // self.frame_rates[name]
+ nb_t[name] = self.sequences[name].nb_images * step[name]
+
+ self.shifts[name] = self.shifts[name] * step[name]
+
+ # Calculate the overall time vector - find the sequence with the
+ # longest total time
+ dt = 1 / self.frame_rate_total
+ for name in self.names:
+ nb_t[name] = self.sequences[name].nb_images * step[name] + self.shifts[name]
+ nb_t = max(nb_t.values())
+ it = range(nb_t)
+ self.time = np.asarray([i * dt for i in it])
+
+ # Calculate indices into every sequence of all time steps.
+ self.index = dict.fromkeys(self.names)
+ for name in self.names:
+ index_cur = [None] * nb_t
+ for frame in range(self.sequences[name].nb_images):
+ ind = step[name] * frame + self.shifts[name]
+ index_cur[ind] = frame
+
+ self.index[name] = index_cur
+
+ def get(self, name: str) -> ImageSequence:
+ """Retrieves an image sequence from its name.
+
+ Args:
+ name: The name (key) of the image sequence to load.
+
+ Returns:
+ The image sequence of the specified name (key).
+ """
+ return self.sequences[name]
+
+ # def get_name(self, index: int) -> str:
+ # """Retrieves the name (key) of the image sequences of the specified index.
+ #
+ # Args:
+ # index: The index of the image sequence to get the name of.
+ #
+ # Returns:
+ # The name (key) of the image sequence of the specified index.
+ # """
+ # return self.names[index]
+ #
+ # def get_names(self) -> List[str]:
+ # """Retrieves the names (keys) of all image sequences.
+ #
+ # Returns:
+ # The names (keys) of all the image sequences
+ # """
+ # return self.names
+
+ def get_path(self, name) -> str:
+ """Retrieves the path of the directory where the image sequence from the specified index are.
+
+ Args:
+ name: The name (key) of the image sequence to load.
+
+ Returns:
+ The path of the directory with image sequence for the specified index
+ """
+
+ return self._paths[self.names.index(name)]
+
+ def get_paths(self) -> List[str]:
+ """Retrieves the paths of the directories where the images of the sequences are.
+
+ Returns:
+ The paths of the directories with the image sequences
+ """
+ return self._paths
+
+ def __getitem__(self, item: Tuple[str, int]) -> Optional[Image]:
+ """Returns the image of sequence name item[0], frame item[1].
+
+ This returns None if no frame is defined at this time step (item[1])
+ for this image sequence name (item[0]).
+
+ Returns:
+ The image of sequence with the name item[0] at the frame
+ item[1].
+ """
+ name, frame = item
+
+ ind = self.index[name][frame]
+ if ind is not None:
+ return self.sequences[name][frame]
+
+ else:
+ return None
+
+ @staticmethod
+ def from_directory(paths: List[str],
+ frame_rates: List[float] or int,
+ names: List[any] = None,
+ shifts: List[int] = None,
+ expr: str = r"\d+(?=\.(tif|png|jpg|bmp))",
+ use_caching: bool = False,
+ check: bool = False,
+ type_image: str = "grey") -> "MultiImageSequence":
+ """Gathers recording files from directory.
+
+ This function sets the files attribute of the instance to a list of
+ subdirectories from the specified directory.
+ In each of these subdirectories, it matches the regular expression
+ and optionally the recording names. If no recording names are
+ specified, the list is in alphabetical order. If recording names
+ are specified, the output is sorted in the same order
+
+ Args:
+ paths: List of paths to directories containing each with a recording sequence.
+ frame_rates: A dictionary of the frame rates of every sequence,
+ with the identifiers as keys or an int if the frame_rate is the same for all.
+ names: A List of names for the sequences. If None, the names of the folder will be used
+ shifts: A dictionary of shifts in number of frames of the start of
+ each sequence with respect to a common time point, with the
+ identifiers as keys. This is used when cameras are running at a
+ different frame rate, but the video sequences do not start at
+ the exact same time instance (as could happen when e.g.,
+ and end trigger is used). For example, a camera running at
+ half the frame rate of another camera may have a shift of one
+ frame w.r.t. the other.
+ expr: Regular expression pattern matching image files.
+ use_caching: Whether or not to use caching for the recordings.
+ check: Whether or not to check all recordings for correctness.
+ type_image: The type of the images ('rgb', 'grey' or 'bw').
+ """
+
+ if names is None:
+ names = [os.path.basename(os.path.normpath(directory)) for directory in paths]
+
+ if type(frame_rates) is int:
+ frame_rates = {name: frame_rates for name in names}
+
+ else:
+ assert len(frame_rates) == len(paths)
+ frame_rates = {file_name: frame_rates[i] for i, file_name in enumerate(names)}
+
+ if shifts is not None:
+ assert len(shifts) == len(paths)
+ shifts = {file_name: shifts[i] for i, file_name in enumerate(names)}
+
+ sequences = {}
+ for i, directory in enumerate(paths):
+ sequences[names[i]] = ImageSequence.from_directory(directory, expr, use_caching=use_caching,
+ check=check, type_image=type_image)
+
+ return MultiImageSequence(sequences, frame_rates, shifts, paths)
+
diff --git a/images/process.py b/images/process.py
deleted file mode 100644
index a643b95e17ccd59e14d9c21805b19a98262cb60a..0000000000000000000000000000000000000000
--- a/images/process.py
+++ /dev/null
@@ -1,3136 +0,0 @@
-import time, shutil, yaml, glob, pickle
-
-from scipy import signal
-import scipy.optimize as optimize
-
-from sympy import solve, symbols
-
-from multiprocessing import Pool
-from functools import partial
-
-import cv2
-from PIL import ImageDraw
-
-# import deeplabcut
-# import tensorflow as tf
-#
-# config = tf.compat.v1.ConfigProto()
-# config.gpu_options.allow_growth = True
-# config.log_device_placement = True
-# sess = tf.compat.v1.Session(config=config)
-
-import matplotlib.pyplot as plt
-
-from cam import calib
-import images.utils as img_utils
-
-# TODO remove specifics dependencies to insect body and wings modules => make more flexible
-from skeleton_fitter.skeleton_modules.insect_body_slim import init_body_params, init_body_bounds, init_body_skeleton3d, \
- build_body_skeleton3d, rotate_body_skeleton3d, translate_body_skeleton3d, estimate_body_parameters_from_skeleton3d
-from skeleton_fitter.skeleton_modules.insect_hybrid_wings_body_slim import init_hybrid_params, init_hybrid_skeleton3d, \
- build_hybrid_skeleton3d
-from skeleton_fitter.skeleton_modules.insect_wings import init_wings_params, init_wing_bounds, init_wings_skeleton_geometry3d_from_csv, \
- build_wing_skeleton3d, rotate_and_translate_wing_skeleton3d, estimate_wing_parameters_from_skeleton3d
-
-from skeleton_fitter.optimiser import optim_fit_body_params, optim_fit_limbs_params
-from skeleton_fitter.utils import reproject_skeleton3d_to2d, angle_from_vectors, length_from_vector
-
-from point_tracker.tracker2d import Tracker2D
-from point_tracker.reconstructor3d import Reconstructor3D
-from skeleton_fitter.cost_functions.fly import *
-from images.read import *
-
-
-def find_char(s, ch):
- return [i for i, ltr in enumerate(s) if ltr == ch]
-
-
-def minimize_perp_distance(x, y, z):
- def model(params, xyz):
- a, b, c, d = params
- x, y, z = xyz
- length_squared = a ** 2 + b ** 2 + c ** 2
- return ((a * x + b * y + c * z + d) ** 2 / length_squared).sum()
-
- def unit_length(params):
- a, b, c, d = params
- return a ** 2 + b ** 2 + c ** 2 - 1
-
- # Initial guess of params using first three points
- p1 = np.squeeze(np.transpose([x[0], y[0], z[0]]))
- p2 = np.squeeze(np.transpose([x[1], y[1], z[1]]))
- p3 = np.squeeze(np.transpose([x[2], y[2], z[2]]))
-
- v1, v2 = p3 - p1, p2 - p1 # These two vectors are in the plane
- cp = np.cross(v1, v2) # the cross product is a vector normal to the plane
- d = np.dot(cp, p3) # This evaluates a * x3 + b * y3 + c * z3 which equals d
- a, b, c = cp
-
- # constrain the vector perpendicular to the plane be of unit length
- cons = ({'type': 'eq', 'fun': unit_length})
- sol = optimize.minimize(model, (a, b, c, d), args=[x, y, z], constraints=cons)
-
- return tuple(sol.x)
-
-
-def filter_interp(y, framerate, cutoff_frequency):
- y_filtered = y.copy()
- is_nan, x = np.isnan(y), lambda z: z.nonzero()[0]
-
- if sum(~is_nan) == 0: return y_filtered # only nans
- if sum(is_nan) > 0: # interpolation needed for filter_high_freq
- y_filtered[is_nan] = np.interp(x(is_nan), x(~is_nan), y_filtered[~is_nan])
-
- w = cutoff_frequency / (framerate / 2) # Normalize the frequency
- b, a = signal.butter(2, w, 'low')
-
- y_filtered = signal.filtfilt(b, a, y_filtered)
- y_filtered[is_nan] = np.nan
-
- return y_filtered
-
-
-def delete_recordings(rec_dates, cam_paths):
- for camn in cam_paths.keys():
-
- if not len(rec_dates[camn]) == 0:
- if not os.path.exists(os.path.join(cam_paths[camn], 'to_delete')):
- os.makedirs(os.path.join(cam_paths[camn], 'to_delete'))
-
- rec_name = 'cam{0}_'.format(camn) + time.strftime('%Y%m%d_%H%M%S', rec_dates[camn])
-
- os.rename(os.path.join(cam_paths[camn], rec_name),
- os.path.join(cam_paths[camn], 'to_delete', rec_name))
- # shutil.rmtree(os.path.join(self.cam_paths[camn], rec_name), ignore_errors=True)
-
- # print('>> recording from {0} has been deleted'.format(time.strftime('%Y%m%d_%H%M%S', rec_dates[camn])))
-
-
-class ImagesProcessing:
- def __init__(self, nb_cam, dlt_path):
-
- self.cam_paths = {1: '', 2: '', 3: ''}
- self.cam_save_paths = {1: '', 2: '', 3: ''}
- self.nb_cam = nb_cam
-
- self.load_dlt_coefs(dlt_path)
-
- self.show_plot = False
-
- self.multiprocessing = True
- self.nb_process = 5
-
- self.main_camn = 1
- self.top_camn = 2
- self.image_format = 'tif'
-
- self.nb_leading_zero = 4
- self.update_leading_zero = False # the number of leading zeros should be 4
-
- self.max_diff_date_s = 25
-
- self.threshold_likelihood = 0.9
- self.threshold_nb_pts_body = 4
- self.threshold_nb_pts_wing = 4
-
- self.cutoff_frequency = 100 # Hz (for Butterworth filter on body params)
- self.framerate = 12500
-
- self.body_labels = ['proboscis_tip', 'proboscis_head_joint', 'head_torso_joint', 'right_wing_hinge',
- 'left_wing_hinge',
- 'torso_abdomen_joint', 'abdomen_middle', 'abdomen_tip']
- self.wing_labels = ['hinge', 'leading_edge_q1', 'leading_edge_q2', 'leading_edge_q3', 'tip', 'trailing_edge_q3',
- 'trailing_edge_q2', 'trailing_edge_q1']
- # self.leg_labels = ['femur_tibia_joint', 'tibia_tarsus_joint', 'tip']
- # self.leg_names = ['front', 'middle', 'back']
-
- self.body_labels_to_keep_cst = ['proboscis_l', 'torso_l', 'torso_r', 'abdomen_l',
- 'proboscis_torso_a', 'torso_abdomen_a', 'ratio_com_torso', 'ratio_proboscis_head']
- self.wing_labels_to_keep_cst = ['span', 'chord', 'aspect_ratio']
- self.keep_init_aspect_ratio = True
-
- self.all_folders_cam, self.all_folders_rec_cam, self.all_files_rec_cam, self.all_dates_rec_cam = {}, {}, {}, {}
- self.skeleton2d, self.skeleton3d = {}, {}
-
- def load_dlt_coefs(self, dlt_path):
- self.dlt_path = dlt_path
- self.dlt_coefs = np.zeros((self.nb_cam, 12))
-
- dlt_csv = np.genfromtxt(dlt_path, delimiter=',', skip_header=0, names=True)
- dlt_csv.dtype.names = ['cam{0}'.format(camn) for camn in range(1, self.nb_cam + 1)]
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- self.dlt_coefs[i] = np.append(np.array(dlt_csv['cam{0}'.format(camn)]), 1)
-
- def init_paths_names(self):
-
- self.nb_cam = int(sorted(self.cam_paths.keys())[-1])
-
- number_rec = [None] * self.nb_cam
- all_files_rec, all_images_rec = {}, {}
- for camn in range(1, self.nb_cam + 1):
- self.all_folders_cam[camn] = sorted(os.listdir(self.cam_paths[camn]), key=lambda x: x[5:20])
- self.all_folders_rec_cam[camn] = [s for s in self.all_folders_cam[camn] if 'cam{0}_'.format(camn) in s]
- number_rec[camn - 1] = len(self.all_folders_rec_cam[camn])
-
- self.all_dates_rec_cam[camn] = {}
- for i, rec_name in enumerate(self.all_folders_rec_cam[camn]):
- self.all_dates_rec_cam[camn][i] = time.strptime(rec_name, 'cam{0}_%Y%m%d_%H%M%S'.format(camn))
-
- if self.update_leading_zero: # Change number of leading zeros (from 6 to 4) if needed
- for i, rec_name in enumerate(self.all_folders_rec_cam[camn]):
- rec_path = os.path.join(self.cam_paths[camn], rec_name)
-
- all_files_rec[camn] = os.listdir(os.path.join(rec_path))
- all_images_rec[camn] = [s for s in all_files_rec[camn] if '.{0}'.format(self.image_format) in s]
-
- if all_images_rec[camn] == []: continue
- old_nb_leading_zeros = all_images_rec[camn][0].find('.{0}'.format(self.image_format)) - len(rec_name)
- all_images_rec[camn] = sorted(all_images_rec[camn], key=lambda x: x[20:20 + old_nb_leading_zeros])
-
- for image_name in all_images_rec[camn]:
- old_nb_leading_zeros = image_name.find('.') - len(rec_name)
- if old_nb_leading_zeros == self.nb_leading_zero: continue
-
- frame = int(image_name[len(rec_name):image_name.find('.{0}'.format(self.image_format))])
- os.rename(os.path.join(rec_path, rec_name + '{0:0{1}d}.{2}'.format(frame, old_nb_leading_zeros,
- self.image_format)),
- os.path.join(rec_path, rec_name + '{0:0{1}d}.{2}'.format(frame, self.nb_leading_zero,
- self.image_format)))
-
- if not all(np.array(number_rec) / number_rec[0]):
- print('ERROR: Different number of recordings have been found for each cameras!')
-
- def gen_rec_names_list(self, main_rec_names):
- rec_names_list = {}
- for camn in range(1, self.nb_cam + 1):
- rec_names_list[camn] = []
-
- for main_rec_name in main_rec_names:
- main_camn = int(main_rec_name[3:4])
- main_rec_date = time.strptime(main_rec_name, 'cam{0}_%Y%m%d_%H%M%S'.format(main_camn))
-
- for camn in range(1, self.nb_cam + 1):
- rec_date = self.nearest_date(self.all_dates_rec_cam[camn], main_rec_date)
-
- if len(rec_date) == 0:
- print('ERROR: Dates were too different from each other! (will delete recording)')
- print(time.strftime('cam #{0} vs. cam{1}_%Y%m%d_%H%M%S'.format(camn, main_camn), main_rec_date))
- continue
-
- rec_names_list[camn].append('cam{0}_'.format(camn) + time.strftime('%Y%m%d_%H%M%S', rec_date))
-
- return rec_names_list
-
- def do_batch(self, fn_name, **kwargs):
- start = time.time()
-
- step_frame = 1 if 'step_frame' not in kwargs.keys() else kwargs['step_frame']
- camn_to_process = list(range(1, self.nb_cam +1) if 'camns' not in kwargs.keys() else kwargs['camns'])
- from_fn_paths = self.get_path_process(kwargs['from_fn_name']) if 'from_fn_name' in kwargs.keys() else self.get_path_process('')
-
- if fn_name in ['autocontrast', 'equalize']: kwargs['factor'] = None;
-
- if fn_name in 'analyse_dlc':
- if 'trainingsetindex' not in kwargs.keys(): kwargs['trainingsetindex'] = 0
- if 'batchsize' not in kwargs.keys(): kwargs['batchsize'] = 2
-
- if fn_name in ['fit_skeleton', 'plot_skeleton']:
- self.body_params_init = init_body_params()
- self.hybrid_params_init = init_hybrid_params()
- self.wings_params_init = init_wings_params(self.body_params_init)
-
- self.body_skeleton3d_init = init_body_skeleton3d(self.body_params_init)
- self.hybrid_skeleton3d_init = init_hybrid_skeleton3d(self.hybrid_params_init)
- self.wings_skeleton3d_init, self.wings_geometry_init = \
- init_wings_skeleton_geometry3d_from_csv(self.wings_params_init, kwargs['csv_path'])
-
- self.body_bounds_init = init_body_bounds()
- self.hybrid_bounds_init = init_body_bounds()
- self.wing_bounds_init = init_wing_bounds(self.body_bounds_init)
-
- self.print_process(fn_name, kwargs)
- process_paths = self.get_path_process(fn_name)
- if process_paths == {1: {}}: process_paths = from_fn_paths
-
- if 'rec_names' in kwargs.keys():
- self.all_folders_rec_cam = kwargs['rec_names'].copy()
- del (kwargs['rec_names'])
-
- if 'yaml_path' in kwargs.keys():
- processes_dict = yaml.load(open(os.path.join(kwargs['yaml_path'])))
-
- main_rec_names = self.all_folders_rec_cam[self.main_camn]
- for main_rec_name in main_rec_names:
- main_rec_date = time.strptime(main_rec_name, 'cam{0}_%Y%m%d_%H%M%S'.format(self.main_camn))
-
- #if main_rec_name[5:13] == '20200229' and int(main_rec_name[-6:]) <= 13915: continue
-
- start_rec = time.time()
- will_delete_rec = False
- rec_names, rec_dates, rec_paths = {}, {}, {}
- images_names, image_paths, image_save_paths, frames = {}, {}, {}, {}
- for camn in range(1, self.nb_cam + 1):
- rec_dates[camn] = self.nearest_date(self.all_dates_rec_cam[camn], main_rec_date)
-
- if len(rec_dates[camn]) == 0:
- print('ERROR: Dates were too different from each other! (will delete recording)')
- print(time.strftime('cam #{0} vs. cam{1}_%Y%m%d_%H%M%S'.format(camn, self.main_camn), main_rec_date))
- will_delete_rec = True
- continue
-
- rec_names[camn] = 'cam{0}_'.format(camn) + time.strftime('%Y%m%d_%H%M%S', rec_dates[camn])
- rec_paths[camn] = os.path.join(self.cam_paths[camn], rec_names[camn])
-
- image_paths[camn] = os.path.join(from_fn_paths[camn], rec_names[camn])
- image_save_paths[camn] = os.path.join(process_paths[camn], rec_names[camn])
-
- files_names = sorted(os.listdir(image_paths[camn]), key=lambda x: x[20:20+self.nb_leading_zero])
- images_names[camn] = [s for s in files_names if '.{0}'.format(self.image_format) in s]
- frames[camn] = sorted([int(s[20:20+self.nb_leading_zero]) for s in images_names[camn]])
- if frames[camn]:
- frames_new = set(frames[camn]) & set(range(1, frames[camn][-1]+1, step_frame))
- images_names[camn] = [image_name for image_name in images_names[camn] if int(image_name[20:20+self.nb_leading_zero]) in frames_new]
- else:
- images_names[camn] = []
- if image_paths[camn] == os.path.join(self.cam_paths[camn], rec_names[camn]): will_delete_rec = True
-
- process_dict = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- if will_delete_rec or camn not in camn_to_process: continue
-
- if 'delete_previous' in kwargs.keys() and kwargs['delete_previous']:
- if os.path.exists(image_save_paths[camn]) and rec_paths[camn] != image_save_paths[camn]:
- shutil.rmtree(image_save_paths[camn], ignore_errors=True)
-
- if not os.path.exists(image_save_paths[camn]): os.makedirs(image_save_paths[camn])
-
- # Read history of processes in a .yaml file
- process_history_yaml_path = os.path.join(image_paths[camn], rec_names[camn] + '-process_history.yaml')
- process_dict[camn], cpt = {}, 1
- if os.path.exists(process_history_yaml_path):
- process_dict[camn] = yaml.load(open(process_history_yaml_path))
-
- if 0 not in process_dict[camn].keys():
- last_frame = int(np.max([frames[i][-1] for i in range(1, self.nb_cam + 1)]))
- missing_frames = {i: list(set(range(1, last_frame +1)) - set(frames[i])) for i in range(1, self.nb_cam + 1)}
- img_size = Image.open(os.path.join(image_paths[camn], images_names[camn][0])).size
- process_dict[camn][0] = {'fn': 'raw', 'camn': camn, 'nb_cam': self.nb_cam, 'rec_names': rec_names,
- 'paths': self.cam_paths, 'process_paths': self.cam_save_paths, 'img_size': img_size,
- 'last_frame': last_frame, 'missing_frames': missing_frames}
-
- for camn in range(1, self.nb_cam + 1):
- if will_delete_rec or camn not in camn_to_process: continue
- else: self.img_size = process_dict[camn][0]['img_size']
-
- cpt = 1
- while cpt in process_dict[camn].keys(): cpt += 1
- process_dict[camn][cpt] = {'fn': fn_name, 'kwargs': kwargs}
-
- if fn_name in ['analyse_dlc', 'load_dlc', 'recon3d', 'stitch', 'multi_processes']: continue
-
- if fn_name == 'crop':
- coord_objs, obj_names = self.interp_2d_obj(rec_paths[camn], image_paths[camn], image_save_paths[camn])
-
- files_names = sorted(os.listdir(rec_paths[camn]), key=lambda x: x[20:20+self.nb_leading_zero])
- images_names[camn] = [s for s in files_names if '.{0}'.format(self.image_format) in s]
- image_paths[camn] = rec_paths[camn]
-
- if fn_name == 'track2d':
- self.track_2d_objects(image_paths[camn], rec_names[camn], image_paths[camn])
-
- elif fn_name == 'save_strobe':
- radius = 50 if 'radius' not in kwargs.keys() else kwargs['radius']
- shape = 'rectangle' if 'shape' not in kwargs.keys() else kwargs['shape']
-
- coord_objs, obj_names = self.interp_2d_obj(rec_paths[camn], image_paths[camn], image_save_paths[camn])
- self.save_stroboscopic_images(coord_objs, image_paths[camn], rec_names[camn], image_save_paths[camn],
- radius, shape, step_frame)
-
- elif fn_name == 'save_avi':
- if os.path.exists(image_paths[camn]):
- framerate = 24 if 'framerate' not in kwargs.keys() else kwargs['framerate']
- lossless = False if 'lossless' not in kwargs.keys() else kwargs['lossless']
-
- self.save_avi(framerate, step_frame, lossless, image_paths[camn], image_save_paths[camn])
-
- elif fn_name == 'delete':
- if rec_names[camn] in kwargs['to_del_names']:
- will_delete_rec = True
-
- else:
- for i, image_name in enumerate(images_names[camn]):
- img = read_image(os.path.join(image_paths[camn], image_name))
- image_save_name = image_name
-
- if fn_name == 'crop':
- for obj_name in obj_names:
- x_pxs, y_pxs = np.array([]), np.array([])
- for x_name in [s for s in coord_objs[obj_name].keys() if 'x_' in s]:
- x_pxs = np.append(x_pxs, coord_objs[obj_name][x_name][i])
- y_pxs = np.append(y_pxs, coord_objs[obj_name][x_name.replace('x_', 'y_')][i])
-
- if not x_pxs.size or not x_pxs.size:
- print('ERROR: x_px and/or y_px is empty!!')
- continue
-
- img = img_utils.crop(img, x_pxs, y_pxs, kwargs['height_crop'], kwargs['width_crop'])
- image_save_name = image_name[:-len(self.image_format) - 1] \
- + '-' + obj_name + '.{0}'.format(self.image_format)
-
- elif fn_name == 'rotate':
- img = img.rotate(kwargs['degrees'])
-
- elif fn_name == 'resize':
- img = img_utils.resize(img, kwargs['resize_ratio'])
-
- elif fn_name in ['enhance_contrast', 'enhance_brightness', 'enhance_sharpness', 'enhance_color',
- 'autocontrast', 'equalize']:
- img = img_utils.enhance(img, fn_name, kwargs['factor'])
-
- elif fn_name == 'save_tiff':
- img_utils.save_tiff(image_name, image_paths[camn], image_save_paths[camn])
-
- if fn_name in ['crop', 'rotate', 'resize', 'sample', 'enhance_contrast', 'enhance_brightness',
- 'enhance_sharpness', 'enhance_color', 'autocontrast', 'equalize']:
- img.save(os.path.join(image_save_paths[camn], image_save_name))
- img.close()
-
- if will_delete_rec:
- delete_recordings(rec_dates, self.cam_paths)
-
- elif fn_name == 'stitch':
- self.stitch_images(images_names, image_paths, image_save_paths)
-
- elif fn_name == 'recon3d':
- pts_csv = {}
- for camn in range(1, self.nb_cam + 1):
- pts_csv[camn] = np.genfromtxt(os.path.join(image_paths[camn], rec_names[camn] + '-2d_points.csv'),
- delimiter=',', skip_header=0, names=True)
-
- self.reconstruct_3d_tracks(pts_csv, rec_names, image_save_paths)
-
- elif fn_name == 'analyse_dlc':
- video_paths = glob.glob(os.path.join(image_paths[self.main_camn], '*.avi'))
-
- if not len(video_paths) == 0:
- deeplabcut.analyze_videos(kwargs['cfg_path'], video_paths, destfolder=image_save_paths[self.main_camn],
- shuffle=kwargs['shuffle'], trainingsetindex=kwargs['trainingsetindex'],
- batchsize=kwargs['batchsize'], save_as_csv=True)
-
- if 'save_avi' in kwargs.keys() and kwargs['save_avi']:
- deeplabcut.create_video_with_all_detections(image_save_paths[self.main_camn], video_paths,
- kwargs['model_name'], destfolder=image_save_paths[self.main_camn])
-
- elif fn_name == 'load_dlc':
- # self.load_dlc_from_csv(kwargs['model_name'], image_paths[self.main_camn], kwargs['tracker_method'])
- # self.load_dlc_from_hdf5(kwargs['model_name'], image_paths[self.main_camn], kwargs['tracker_method'])
- self.load_dlc_from_full_pickle(kwargs['model_name'], image_paths[self.main_camn])
-
- #self.test_unwrap()
- self.unscramble_views2d_dlc(process_dict)
- self.reverse_processes_2d_points(process_dict)
- self.save2d_dlc(image_save_paths)
-
- #self.unscramble_sides2d_and_recon3d_dlc()
- self.recon3d_dlc()
- self.save3d_dlc(image_save_paths)
-
- elif fn_name == 'fit_skeleton':
- csv_paths = glob.glob(os.path.join(image_paths[self.main_camn], '*-3d_points.csv'))
- if not len(csv_paths) == 0:
- self.load_skeleton3d_from_csv(kwargs['model_name'], image_paths)
- self.load_skeleton2d_from_csv(kwargs['model_name'], image_paths)
-
- if 'angles_to_correct' not in kwargs.keys(): kwargs['angles_to_correct'] = []
- self.optim_fit_skeleton(kwargs['animal_name'], kwargs['body_param_names'], kwargs['wing_param_names'], kwargs['res_method'],
- kwargs['opt_method'], rec_paths, image_save_paths, kwargs['angles_to_correct'])
-
- elif fn_name == 'plot_skeleton':
- csv_paths = glob.glob(os.path.join(image_paths[self.main_camn], '*-2d_points.csv'))
- if not len(csv_paths) == 0:
- self.load_skeleton2d_from_csv(kwargs['model_name'], image_paths)
- body_params, wings_params = self.load_params_from_csv(image_paths[self.main_camn])
-
- self.plot_skeleton2d(body_params, wings_params, rec_paths, step_frame,
- save_images=True, save_paths=image_save_paths)
-
- if 'save_avi' in kwargs.keys() and kwargs['save_avi']:
- framerate = 24 if 'framerate' not in kwargs.keys() else kwargs['framerate']
- lossless = False if 'lossless' not in kwargs.keys() else kwargs['lossless']
- self.save_avi(framerate, step_frame, lossless, image_save_paths[self.main_camn], image_save_paths[self.main_camn])
-
- elif fn_name == 'multi_processes':
- print('> Processing {0}...'.format(main_rec_name))
- multi_processes_dict = {}
- for camn in range(1, self.nb_cam + 1):
- multi_processes_dict[camn] = processes_dict.copy()
- multi_processes_dict[camn][0] = process_dict[camn][0]
- self.multi_processes(multi_processes_dict, images_names, frames, rec_names)
-
- print('> {0} as been processed (time elapsed: {1:.4f} s)'.format(main_rec_name, time.time() - start_rec))
- continue
-
- # Save history
- for camn in range(1, self.nb_cam + 1):
- if will_delete_rec or camn not in camn_to_process: continue
-
- with open(os.path.join(process_paths[camn], rec_names[camn],
- rec_names[camn] + '-process_history.yaml'), 'w') as yaml_file:
- yaml.safe_dump(process_dict[camn], yaml_file, encoding='utf-8', allow_unicode=True)
-
- print('> {0} as been processed (time elapsed: {1:.4f} s)'.format(main_rec_name, time.time() - start_rec))
-
- self.init_paths_names()
- print('> Full batch as been processed (total time elapsed: {0:.4f} s)'.format(time.time() - start))
-
- def print_process(self, fn_name, kwargs):
- print('____________________________')
- if fn_name == 'multi_processes':
- print('> Starting batch images: Running multiple processes')
- elif fn_name == 'sample':
- print('> Starting batch images: Copying samples')
- elif fn_name == 'resize':
- print('> Starting batch images: Up/Down-sizing images (ratio: {0})'.format(kwargs['resize_ratio']))
- elif fn_name == 'enhance_contrast':
- print('> Starting batch images: Enhancing contrast of images (factor: {0})'.format(kwargs['factor']))
- elif fn_name == 'enhance_brightness':
- print('> Starting batch images: Enhancing brightness of images (factor: {0})'.format(kwargs['factor']))
- elif fn_name == 'enhance_sharpness':
- print('> Starting batch images: Enhancing sharpness of images (factor: {0})'.format(kwargs['factor']))
- elif fn_name == 'enhance_color':
- print('> Starting batch images: Enhancing color of images (factor: {0})'.format(kwargs['factor']))
- elif fn_name == 'autocontrast':
- print('> Starting batch images: Auto-Enhancing contrast of images')
- elif fn_name == 'equalize':
- print('> Starting batch images: Equalizing of images')
- elif fn_name == 'crop':
- print('> Starting batch images: Cropping images (height: {0}, width: {1})'.format(kwargs['height_crop'], kwargs['width_crop']))
- elif fn_name == 'rotate':
- print('> Starting batch images: Rotating cameras views {0} at {1} degrees '.format(kwargs['camns'], kwargs['degrees']))
- elif fn_name == 'stitch':
- print('> Starting batch images: Stitching all cameras views together')
- elif fn_name == 'track2d':
- print('> Starting batch images: Tracking 2d coordinates of blobs')
- elif fn_name == 'save_strobe':
- print('> Starting batch images: Generating stroboscopic images from 2d coordinates')
- elif fn_name == 'recon3d':
- print('> Starting batch images: Reconstructing 3d tracks from 2d coordinates')
- elif fn_name == 'save_tiff':
- print('> Starting batch images: Saving recordings as 8 bits .tiff')
- elif fn_name == 'save_avi':
- print('> Starting batch images: Saving recordings as .avi')
- elif fn_name == 'analyse_dlc':
- print('> Starting batch images: Analysing videos with DeepLabCut (model: {0})'.format(kwargs['model_name']))
- elif fn_name == 'load_dlc':
- print('> Starting batch images: Load points from DeepLabCut (unscramble, reconstruct3d ... - model: {0})'.format(kwargs['model_name']))
- elif fn_name == 'fit_skeleton':
- print('> Starting batch images: Optimizing fit btw skeletons to get transformation/rotation parameters')
- elif fn_name == 'plot_skeleton':
- print('> Starting batch images: Plot (or save) skeleton on raw images')
- elif fn_name == 'delete':
- print('> Starting batch images: Deleting following recordings: {0}'.format(kwargs['to_del_names']))
-
- def get_path_process(self, fn_name):
- process_paths = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- if fn_name == '':
- process_paths[camn] = self.cam_paths[camn]
- elif fn_name == 'crop':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Cropped')
- elif fn_name == 'sample':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Sample')
- elif fn_name == 'resize':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Resized')
- elif fn_name in ['enhance_contrast', 'enhance_brightness', 'enhance_sharpness', 'enhance_color',
- 'autocontrast', 'equalize']:
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Enhanced')
- elif fn_name == 'save_tiff':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Tiff')
- elif fn_name == 'save_avi':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Avi')
- elif fn_name == 'save_strobe':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Strobe')
- elif fn_name == 'stitch':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Stitched')
- elif fn_name in ['analyse_dlc', 'load_dlc']:
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_DLC')
- elif fn_name == 'plot_skeleton':
- process_paths[camn] = os.path.join(self.cam_save_paths[camn], '_Plots')
- elif fn_name in ['rotate', 'delete', 'track2d', 'recon3d', 'multi_processes', 'fit_skeleton']:
- break
- else:
- print('WARN: incorrect function name (fn_name = {0})'.format(fn_name))
- exit()
-
- return process_paths
-
- def multi_processes(self, multi_processes_dict, all_images_names, all_frames, rec_names):
- multi_processes_nums = list(multi_processes_dict[1].keys())
- rec_paths = self.get_path_process('')
-
- process_dict = {}
- for camn in range(1, self.nb_cam + 1): process_dict[camn] = {0: multi_processes_dict[camn][0]}
-
- self.img_size = multi_processes_dict[1][0]['img_size']
-
- img, obj_names = {}, []
- for multi_processes_num in multi_processes_nums[1:]:
- start = time.time()
- fn_name = multi_processes_dict[1][multi_processes_num]['fn']
- kwargs = multi_processes_dict[1][multi_processes_num]['kwargs']
- step_frame = 1 if 'step_frame' not in kwargs.keys() else kwargs['step_frame']
- camn_to_process = list(range(1, self.nb_cam + 1) if 'camns' not in kwargs.keys() else kwargs['camns'])
-
- from_fn_paths = self.get_path_process(kwargs['from_fn_name']) if 'from_fn_name' in kwargs.keys() else self.get_path_process('')
- process_paths = self.get_path_process(fn_name)
- if process_paths == {1: {}}: process_paths = from_fn_paths
-
- images_names, image_rec_paths, image_paths, image_save_paths = {}, {}, {}, {}
- for camn in range(1, self.nb_cam + 1):
- if all_frames[camn]:
- frames_new = set(all_frames[camn]) & set(range(1, all_frames[camn][-1], step_frame))
- images_names[camn] = [s for s in all_images_names[camn] if int(s[20:20+self.nb_leading_zero]) in frames_new]
- else:
- images_names[camn] = []
-
- image_rec_paths[camn] = os.path.join(rec_paths[camn], rec_names[camn])
- image_paths[camn] = os.path.join(from_fn_paths[camn], rec_names[camn])
- image_save_paths[camn] = os.path.join(process_paths[camn], rec_names[camn])
-
- if 'delete_previous' in kwargs.keys() and kwargs['delete_previous']:
- if os.path.exists(image_save_paths[camn]) and rec_paths[camn] != image_save_paths[camn]:
- shutil.rmtree(image_save_paths[camn], ignore_errors=True)
-
- for camn in range(1, self.nb_cam +1):
- process_history_yaml_path = os.path.join(image_paths[camn], rec_names[camn] + '-process_history.yaml')
- if os.path.exists(process_history_yaml_path):
- process_dict[camn] = yaml.load(open(process_history_yaml_path))
-
- if not obj_names or 'obj0' in obj_names:
- file_names = sorted(os.listdir(image_paths[camn]), key=lambda x: x[20:20 + self.nb_leading_zero])
- csv_names = [s for s in file_names if '-2d_points.csv' in s and '-obj' in s]
-
- if csv_names:
- obj_names = np.unique([s[find_char(s, '-')[0] + 1:find_char(s, '-')[1]] for s in csv_names])
- if 'obj0' in img[camn].keys(): img[camn] = {}
- else:
- obj_names = ['obj0']
-
- if not os.path.exists(image_save_paths[camn]): os.makedirs(image_save_paths[camn])
- if not camn in camn_to_process: continue
- if fn_name in ['recon3d', 'stitch']: continue
-
- if fn_name == 'crop':
- coord_objs, obj_names = self.interp_2d_obj(image_rec_paths[camn], image_paths[camn], image_save_paths[camn])
-
- elif fn_name == 'track2d':
- self.track_2d_objects(image_paths[camn], rec_names[camn], image_save_paths[camn])
- continue
-
- elif fn_name == 'save_strobe':
- radius = 50 if 'radius' not in kwargs.keys() else kwargs['radius']
- shape = 'rectangle' if 'shape' not in kwargs.keys() else kwargs['shape']
-
- coord_objs, _ = self.interp_2d_obj(image_rec_paths[camn], image_paths[camn], image_save_paths[camn])
- self.save_stroboscopic_images(coord_objs, image_paths[camn], rec_names[camn], image_save_paths[camn],
- radius, shape, step_frame)
- continue
-
- elif fn_name == 'save_avi':
- framerate = 24 if 'framerate' not in kwargs.keys() else kwargs['framerate']
- lossless = False if 'lossless' not in kwargs.keys() else kwargs['lossless']
-
- if camn not in img.keys(): img[camn] = {}
-
- for obj_name in obj_names:
- if obj_name not in img[camn].keys():
- img[camn][obj_name] = {}
-
- #if fn_name not in ['track2d', 'recon3d'] \
- # and obj_name == 'obj0' and 'from_fn_name' in kwargs.keys(): break
-
- if fn_name == 'save_avi':
- if 'obj0' == obj_name: save_name = rec_names[camn] + '.avi'
- else: save_name = rec_names[camn] + '-' + obj_name + '.avi'
-
- if len(np.shape(np.asarray(img[camn][obj_name][0]))) != 2: break
- height_img, width_img = np.shape(np.asarray(img[camn][obj_name][0]))
-
- codec = cv2.VideoWriter_fourcc(*'HFYU') if lossless else cv2.VideoWriter_fourcc(*'MP42')
- writer = cv2.VideoWriter(os.path.join(image_save_paths[camn], save_name), codec, framerate, (width_img, height_img))
- continue
-
- for i, image_name in enumerate(images_names[camn]):
- if i not in img[camn][obj_name].keys():
- img[camn][obj_name][i] = read_image(os.path.join(image_rec_paths[camn], image_name))
-
- if not img[camn][obj_name][i]: continue
-
- if fn_name == 'crop':
- x_pxs, y_pxs = np.array([]), np.array([])
- for x_name in [s for s in coord_objs[obj_name].keys() if 'x_' in s]:
- x_pxs = np.append(x_pxs, coord_objs[obj_name][x_name][i])
- y_pxs = np.append(y_pxs, coord_objs[obj_name][x_name.replace('x_', 'y_')][i])
-
- if not x_pxs.size or not x_pxs.size:
- print('ERROR: x_px and/or y_px is empty!!')
- continue
-
- img[camn][obj_name][i] = \
- img_utils.crop(img[camn][obj_name][i], x_pxs, y_pxs, kwargs['height_crop'], kwargs['width_crop'])
-
- elif fn_name == 'rotate':
- img[camn][obj_name][i] = img[camn][obj_name][i].rotate(kwargs['degrees'])
-
- elif fn_name == 'resize':
- img[camn][obj_name][i] = img_utils.resize(img[camn][obj_name][i], kwargs['resize_ratio'])
-
- elif fn_name in ['enhance_contrast', 'enhance_brightness', 'enhance_sharpness', 'enhance_color',
- 'autocontrast', 'equalize']:
- img[camn][obj_name][i] = img_utils.enhance(img[camn][obj_name][i], fn_name, kwargs['factor'])
-
- elif fn_name == 'save_avi':
- writer.write(cv2.cvtColor(np.asarray(img[camn][obj_name][i]), cv2.COLOR_GRAY2BGR))
-
- if fn_name == 'save_avi': writer.release()
-
- if 'step_frame' in kwargs.keys() or multi_processes_num == multi_processes_nums[-1]: # Will save images
- if fn_name in ['crop', 'rotate', 'resize', 'sample', 'enhance_contrast', 'enhance_brightness',
- 'enhance_sharpness', 'enhance_color', 'autocontrast', 'equalize']:
- for i, image_name in enumerate(images_names[camn]):
- if 'obj0' == obj_name: image_save_name = image_name
- else: image_save_name = image_name[:-len(self.image_format) - 1] + '-' + obj_name \
- + '.{0}'.format(self.image_format)
-
- img[camn][obj_name][i].save(os.path.join(image_save_paths[camn], image_save_name))
- #img[camn][obj_name][i] = []
-
- if fn_name == 'recon3d':
- pts_csv = {}
- for camn in range(1, self.nb_cam + 1):
- pts_csv[camn] = np.genfromtxt(os.path.join(image_paths[camn], rec_names[camn] + '-2d_points.csv'),
- delimiter=',', skip_header=0, names=True)
-
- self.reconstruct_3d_tracks(pts_csv, rec_names, image_save_paths)
-
- elif fn_name == 'stitch':
- ref_images, refs_images = set(), {}
- for camn in range(1, self.nb_cam + 1):
- refs_images[camn] = [image_name[20:-(len(self.image_format) + 1)] for image_name in images_names[camn]]
- ref_images = ref_images & set(refs_images[camn]) if len(ref_images) > 0 else set(refs_images[camn])
-
- ref_images = np.sort(list(ref_images))
-
- index_i = {}
- for camn in range(1, self.nb_cam + 1):
- index_i[camn] = [refs_images[camn].index(ref_image) for ref_image in ref_images]
-
- for obj_name in obj_names:
- #if obj_name == 'obj0' and 'from_fn_name' in kwargs.keys(): break
-
- for camn in range(1, self.nb_cam + 1):
- if obj_name not in img[camn].keys():
- img[camn][obj_name] = {}
-
- for i, _ in enumerate(index_i[self.main_camn]):
- imgs = {}
- for camn in range(1, self.nb_cam + 1):
- if i not in img[camn][obj_name].keys():
- img[camn][obj_name][i] = read_image(os.path.join(image_rec_paths[camn], images_names[camn][i]))
-
- imgs[camn] = img[camn][obj_name][i]
- img[camn][obj_name][i] = []
-
- img[self.main_camn][obj_name][i] = img_utils.stitch_images(imgs, self.nb_cam)
-
- if multi_processes_num == multi_processes_nums[-1]:
- for i, image_name in enumerate(images_names[camn]):
- if 'obj0' == obj_name: image_save_name = image_name
- else: image_save_name = image_name[:-len(self.image_format) - 1] + '-' + obj_name + '.{0}'.format(self.image_format)
- img[self.main_camn][obj_name][i].save(os.path.join(image_save_paths[self.main_camn], image_save_name))
- img[self.main_camn][obj_name][i].close()
-
- for camn in range(1, self.nb_cam + 1):
- with open(os.path.join(process_paths[camn], rec_names[camn], rec_names[camn] + '-process_history.yaml'), 'w') as yaml_file:
- process_dict[camn][multi_processes_num] = multi_processes_dict[camn][multi_processes_num]
-
- yaml.safe_dump(process_dict[camn], yaml_file, encoding='utf-8', allow_unicode=True)
-
- print('>> fn {0} as been performed (time elapsed: {1:.4f} s)'.format(fn_name, time.time() - start))
-
- def save_avi(self, framerate, step_frame, lossless, image_path, save_path):
- all_files_names = sorted(os.listdir(image_path), key=lambda x: x[20:20+self.nb_leading_zero])
- all_image_names = [s for s in all_files_names if '.{0}'.format(self.image_format) in s]
- obj_names = np.unique([s[s.find('-')+1:-(len(self.image_format)+1)] for s in all_image_names])
-
- for obj_name in obj_names:
- image_names = sorted([i for i in all_image_names if obj_name in i], key=lambda x: x[20:20+self.nb_leading_zero])
- save_name = image_names[0][:20] + image_names[0][20+self.nb_leading_zero:-(len(self.image_format)+1)] + '.avi'
-
- img = cv2.imread(os.path.join(image_path, image_names[0]), cv2.IMREAD_UNCHANGED)
- is_rgb = len(img.shape) == 3
- if is_rgb: height_img, width_img, _ = img.shape
- else: height_img, width_img = img.shape
-
- # commun: 'MP42', 'XVID' or lossless: 'HFYU'
- codec = cv2.VideoWriter_fourcc(*'HFYU') if lossless else cv2.VideoWriter_fourcc(*'MP42')
- writer = cv2.VideoWriter(os.path.join(save_path, save_name), codec, framerate, (width_img, height_img,))
-
- for i in np.arange(0, len(image_names), step_frame):
- img_16bit = cv2.imread(os.path.join(image_path, image_names[i]), cv2.IMREAD_UNCHANGED)
- if is_rgb: writer.write(img_16bit)
- else: writer.write(cv2.cvtColor(img_16bit, cv2.COLOR_GRAY2BGR))
- #cv2.imshow(obj_name, img_16bit)
- #cv2.waitKey(1)
-
- writer.release()
-
- def track_2d_objects(self, image_path, save_name, save_path):
-
- tracker = Tracker2D()
- tracker.initialize()
- tracker.load_images(image_path, self.image_format)
-
- tracker.do_tracking()
-
- # tracker.show_plot = self.show_plot
- if self.show_plot:
- all_files = sorted(os.listdir(image_path), key=lambda x: x[20:20+self.nb_leading_zero])
- all_images = [s for s in all_files if '.{0}'.format(self.image_format) in s]
-
- fig = plt.figure()
- ax = fig.add_subplot(111)
- plt.imshow(Image.open(os.path.join(image_path, all_images[0])))
- ax.scatter(tracker.points['x'], tracker.points['y'])
- ax.set_xlabel('x (m)')
- ax.set_ylabel('y (m)')
- plt.show()
-
- tracker.save_csv(save_name, save_path)
-
- def save_stroboscopic_images(self, coord_objs, image_path, save_name, save_path, radius, shape, step_frame):
-
- if len(coord_objs.keys()) == 0:
- print('WARM: coord_objs is empty!')
- return
-
- tracker = Tracker2D()
-
- tracker.dist2Threshold = 7
- tracker.initialize()
- tracker.load_images(image_path, self.image_format)
-
- for obj_name in coord_objs.keys():
- tracker.points = {'x': np.array(coord_objs[obj_name]['x_px']), 'y': np.array(coord_objs[obj_name]['y_px']),
- 'frame': np.array(coord_objs[obj_name]['frame']),
- 'area': np.ones(np.size(coord_objs[obj_name]['x_px'])) * np.nan}
-
- tracker.gen_stroboscopic_image(radius, shape, step_frame, flip_time=True)
- tracker.save_stroboscopic_image(save_name + '{0:0{1}d}'.format(1, self.nb_leading_zero) + '-' + obj_name, save_path)
-
- def reconstruct_3d_points(self, pts_csv, save_names, save_paths, save=True):
- reconstructor = Reconstructor3D(self.nb_cam, self.img_size)
- reconstructor.dlt_coefs = self.dlt_coefs
-
- xs_coord = np.array([pts_csv[camn]['x_px'] for camn in range(1, self.nb_cam + 1)])
- ys_coord = np.array([pts_csv[camn]['y_px'] for camn in range(1, self.nb_cam + 1)])
- frames_coord = np.array([pts_csv[camn]['frame'] for camn in range(1, self.nb_cam + 1)])
-
- xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
- reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
-
- save_names_list = [save_names[camn] for camn in range(1, self.nb_cam + 1)]
- save_paths_list = [save_paths[camn] for camn in range(1, self.nb_cam + 1)]
- if save: reconstructor.save_points_csv(save_names_list, save_paths_list, save_type='all_frames')
-
- return xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses
-
- def reconstruct_3d_tracks(self, pts_csv, save_names, save_paths, save=True):
- reconstructor = Reconstructor3D(self.nb_cam, self.img_size)
- reconstructor.dlt_coefs = self.dlt_coefs
-
- xs_coord = np.array([pts_csv[camn]['x_px'] for camn in range(1, self.nb_cam + 1)])
- ys_coord = np.array([pts_csv[camn]['y_px'] for camn in range(1, self.nb_cam + 1)])
- frames_coord = np.array([pts_csv[camn]['frame'] for camn in range(1, self.nb_cam + 1)])
-
- xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts, _, = \
- reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
-
- tracks_dict = reconstructor.recon_tracks(xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts)
-
- if not xs_pose or not tracks_dict['obj']: return
-
- save_names_list = [save_names[camn] for camn in range(1, self.nb_cam + 1)]
- save_paths_list = [save_paths[camn] for camn in range(1, self.nb_cam + 1)]
- if save: reconstructor.save_tracks_csv(save_names_list, save_paths_list)
-
- if self.show_plot:
- xs_pose, ys_pose, zs_pose = np.concatenate(xs_pose), np.concatenate(ys_pose), np.concatenate(zs_pose)
-
- fig = plt.figure()
- ax = fig.add_subplot(111, projection='3d')
- ax.scatter(xs_pose, ys_pose, zs_pose, marker='+')
- for num_obj in tracks_dict['obj'].keys():
- ax.scatter(tracks_dict['obj'][num_obj]['x'], tracks_dict['obj'][num_obj]['y'],
- tracks_dict['obj'][num_obj]['z'], marker='o')
- ax.set_xlabel('x (m)')
- ax.set_ylabel('y (m)')
- ax.set_zlabel('z (m)')
- plt.show()
-
- return tracks_dict
-
- def load_dlc_from_csv(self, model_name, rec_path, tracker_method):
- process_dict = yaml.load(open(glob.glob(os.path.join(rec_path, '*-process_history.yaml'))[0]))
- frames = list(range(1, process_dict[0]['last_frame']))
- missing_frames = list(np.unique(np.concatenate([process_dict[0]['missing_frames'][camn] for camn in range(1, self.nb_cam + 1)])))
- [missing_frames.remove(missing_frame) for missing_frame in list(set(missing_frames) - set(frames))]
- [frames.remove(missing_frame) for missing_frame in missing_frames]
-
- if tracker_method == 'skeleton': tracker_method_str = '_sk'
- elif tracker_method == 'box': tracker_method_str = '_bk'
-
- dlc_csv_paths = glob.glob(os.path.join(rec_path, '*' + model_name + tracker_method_str + '.csv'))
-
- obj_names = []
- for dlc_csv_path in dlc_csv_paths: # go throught the various obj
- obj_names.append(dlc_csv_path[len(rec_path) + 1:dlc_csv_path.index(model_name)])
-
- self.skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
- for i, obj_name in enumerate(obj_names):
- df = pd.read_csv(dlc_csv_paths[i], header=[1, 2])
- label_names = np.unique([s[0][:-2] for s in df.keys()[1:]])
- new_label_names = [label.replace(' ', '_') for label in label_names]
-
- self.skeleton2d[obj_name] = {'frames': frames, 'label_names': new_label_names}
- for j, label in enumerate(label_names):
- new_label = new_label_names[j]
-
- self.skeleton2d[obj_name][new_label] = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- label_key = label + ' {0}'.format(camn)
-
- self.skeleton2d[obj_name][new_label][camn] = {'x': np.array(df[(label_key, 'x')]),
- 'y': np.array(df[(label_key, 'y')]),
- 'likelihood': np.array(df[(label_key, 'likelihood')])}
-
- def load_dlc_from_full_pickle(self, model_name, rec_path):
- process_dict = yaml.load(open(glob.glob(os.path.join(rec_path, '*-process_history.yaml'))[0]))
- frames = list(range(1, process_dict[0]['last_frame']))
- missing_frames = list(np.unique(np.concatenate([process_dict[0]['missing_frames'][camn] for camn in range(1, self.nb_cam + 1)])))
- [missing_frames.remove(missing_frame) for missing_frame in list(set(missing_frames) - set(frames))]
- [frames.remove(missing_frame) for missing_frame in missing_frames]
-
- dlc_pickle_paths = glob.glob(os.path.join(rec_path, '*' + model_name + '_full.pickle'))
-
- obj_names = []
- for dlc_pickle_path in dlc_pickle_paths: # go throught the various obj
- obj_names.append(dlc_pickle_path[len(rec_path) + 22:dlc_pickle_path.index(model_name)])
-
- self.skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
- for obj_name in obj_names:
- with open(dlc_pickle_paths[0], "rb") as file:
- full_dict = pickle.load(file)
-
- header = full_dict.pop("metadata")
- label_names = np.unique([s[:-2] for s in header["all_joints_names"]])
- new_label_names = [label.replace(' ', '_') for label in label_names]
-
- self.skeleton2d[obj_name] = {'frames': frames, 'label_names': new_label_names}
- for i, label in enumerate(label_names):
- new_label = new_label_names[i]
-
- self.skeleton2d[obj_name][new_label] = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- self.skeleton2d[obj_name][new_label][camn] = {'x': np.nan * np.ones(len(frames)),
- 'y': np.nan * np.ones(len(frames)),
- 'likelihood': np.nan * np.ones(len(frames))}
-
- for frame_name in full_dict.keys():
- try:
- frame_in = frames.index(int(frame_name[-4:]) + 1)
- except ValueError:
- continue
-
- for i, label in enumerate(label_names):
- new_label = new_label_names[i]
-
- for camn in range(1, self.nb_cam + 1):
- try: label_in = header["all_joints_names"].index(label + ' {0}'.format(camn))
- except ValueError: continue
-
- if len(full_dict[frame_name]['confidence'][label_in]) == 0: continue
- best_in = np.argmax(full_dict[frame_name]['confidence'][label_in])
-
- self.skeleton2d[obj_name][new_label][camn]['x'][frame_in] = \
- full_dict[frame_name]['coordinates'][0][label_in][best_in][0]
- self.skeleton2d[obj_name][new_label][camn]['y'][frame_in] = \
- full_dict[frame_name]['coordinates'][0][label_in][best_in][1]
- self.skeleton2d[obj_name][new_label][camn]['likelihood'][frame_in] = \
- full_dict[frame_name]['confidence'][label_in][best_in]
-
- def load_dlc_from_hdf5(self, model_name, rec_path, tracker_method):
- process_dict = yaml.load(open(glob.glob(os.path.join(rec_path, '*-process_history.yaml'))[0]))
- frames = list(range(1, process_dict[0]['last_frame']))
- missing_frames = np.unique(
- [np.squeeze(process_dict[0]['missing_frames'][camn]) for camn in range(1, self.nb_cam + 1)])
- [frames.remove(missing_frame) for missing_frame in missing_frames]
-
- if tracker_method == 'skeleton': tracker_method_str = '_sk'
- elif tracker_method == 'box': tracker_method_str = '_bk'
-
- dlc_hdf5_paths = glob.glob(os.path.join(rec_path, '*' + model_name + tracker_method_str + '.h5'))
-
- obj_names = []
- for dlc_hdf5_path in dlc_hdf5_paths: # go throught the various obj
- obj_names.append(dlc_hdf5_path[len(rec_path) + 22:dlc_hdf5_path.index(model_name)])
-
- self.skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
- for i, obj_name in enumerate(obj_names):
- hdf5_dict = pd.read_hdf(dlc_hdf5_paths[i], 'df_with_missing')
- # hdf5_dict.to_csv(os.path.join(save_path, hdf5_name.split(".h5")[0] + '.csv'))
-
- # network_name = hdf5_dict.columns[0][0]
- hdf5_dict.columns = hdf5_dict.columns.droplevel([0, 1]) # remove first two lines of header (network_name, individual)
-
- label_names = np.unique([s[:-2] for s in hdf5_dict.columns.droplevel(1)])
- new_label_names = [label.replace(' ', '_') for label in label_names]
- # frames = hdf5_dict.index
-
- self.skeleton2d[obj_name] = {'frames': frames, 'label_names': new_label_names}
- for i, label in enumerate(label_names):
- new_label = new_label_names[i]
-
- self.skeleton2d[obj_name][new_label] = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- label_key = label + ' {0}'.format(camn)
-
- self.skeleton2d[obj_name][new_label][camn] = {'x': np.array(hdf5_dict[(label_key, 'x')]),
- 'y': np.array(hdf5_dict[(label_key, 'y')]),
- 'likelihood': np.array(hdf5_dict[(label_key, 'likelihood')])}
-
- # self.skeleton2d[obj_name][new_label][camn] = [[1]*3 for i in range(len(hdf5_dict[(label_key, 'x')]))]
- # # self.skeleton2d[obj_name][new_label][camn] = np.ones((len(hdf5_dict[(label_key, 'x')]), 3))
- # self.skeleton2d[obj_name][new_label][camn][:, 0] = np.array(hdf5_dict[(label_key, 'x')])
- # self.skeleton2d[obj_name][new_label][camn][:, 1] = np.array(hdf5_dict[(label_key, 'y')])
- # self.skeleton2d[obj_name][new_label][camn][:, 2] = np.array(hdf5_dict[(label_key, 'likelihood')])
-
- def load_skeleton2d_from_csv(self, model_name, paths):
- csv_paths = glob.glob(os.path.join(paths[self.main_camn], '*-2d_points.csv'))
- obj_names = np.unique([csv_path[len(paths[self.main_camn]) + 1:csv_path.index('-')] for csv_path in csv_paths])
-
- self.skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
- for obj_name in obj_names:
- label_names = np.unique([csv_path[csv_path.index('-') + 1:csv_path.index('-2d_points.csv')] for csv_path in csv_paths])
-
- self.skeleton2d[obj_name] = {'frames': [], 'label_names': label_names}
- for label in label_names:
- self.skeleton2d[obj_name][label] = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- pts_csv = np.genfromtxt(os.path.join(paths[camn], obj_name + '-' + label + '-2d_points.csv'),
- delimiter=',', skip_header=0, names=True)
-
- self.skeleton2d[obj_name]['frames'] = \
- np.unique(np.concatenate((self.skeleton2d[obj_name]['frames'], pts_csv['frame'])))
- self.skeleton2d[obj_name][label][camn] = \
- {'x': pts_csv['x_px'], 'y': pts_csv['y_px'], 'likelihood': pts_csv['likelihood']}
-
- def load_skeleton3d_from_csv(self, model_name, paths):
- csv_paths = glob.glob(os.path.join(paths[self.main_camn], '*-3d_points.csv'))
- obj_names = np.unique([csv_path[len(paths[self.main_camn]) + 1:csv_path.index('-')] for csv_path in csv_paths])
-
- if len(csv_paths) == 0:
- raise ValueError("Couldn't find any *-3d_points.csv file (path: {0})".format(os.path.join(paths[self.main_camn])))
-
- self.skeleton3d = {'obj_names': obj_names, 'model_name': model_name}
- for obj_name in obj_names:
- obj_csv_paths = [csv_path for csv_path in csv_paths if obj_name in csv_path]
- label_names = np.unique([csv_path[csv_path.index(obj_name + '-') + len(obj_name) +1:
- csv_path.index('-3d_points.csv')] for csv_path in obj_csv_paths])
-
- self.skeleton3d[obj_name] = {'frames': [], 'label_names': label_names}
- for label in label_names:
- self.skeleton3d[obj_name][label] = {1: {}}
- for camn in range(1, self.nb_cam + 1):
- pts_csv = np.genfromtxt(os.path.join(paths[camn], obj_name + '-' + label + '-3d_points.csv'),
- delimiter=',', skip_header=0, names=True)
-
- self.skeleton3d[obj_name]['frames'] = \
- np.unique(np.concatenate((self.skeleton3d[obj_name]['frames'], pts_csv['frame'])))
- self.skeleton3d[obj_name][label] = {'x': pts_csv['x'], 'y': pts_csv['y'], 'z': pts_csv['z']}
-
- def load_params_from_csv(self, path):
- obj_names = self.skeleton2d['obj_names']
-
- body_params, wings_params = {}, {}
- for obj_name in obj_names:
- body_params[obj_name], wings_params[obj_name] = {}, {}
- params_csv = np.genfromtxt(os.path.join(path, 'skeleton_parameters-' + obj_name + '.csv'),
- delimiter=',', skip_header=0, names=True)
-
- for body_param_name in self.body_params_init.keys():
- body_params[obj_name][body_param_name] = params_csv[body_param_name]
-
- for side in ['right', 'left']:
- wings_params[obj_name][side] = {}
- for wing_param_name in self.wings_params_init[side].keys():
- if wing_param_name in self.body_params_init.keys():
- wings_params[obj_name][side][wing_param_name] = params_csv[wing_param_name]
- else:
- wings_params[obj_name][side][wing_param_name] = params_csv[wing_param_name + '_' + side]
-
- return body_params, wings_params
-
- def unscramble_views2d_dlc(self, process_dict):
- # Run on unconverted 2d coords from dlc
- # shared_axis: 0 if the combination of cam share the x axis (top-left to top-right)
- # or 1 if they share the y axis (top-left to bottom left)
-
- process_names = [process_dict[1][process_num]['fn'] for process_num in process_dict[1].keys()]
- crop_num = process_names.index('crop')
- crop_num = crop_num[-1] if isinstance(crop_num, list) else crop_num
- width_crop = process_dict[1][crop_num]['kwargs']['width_crop']
- height_crop = process_dict[1][crop_num]['kwargs']['height_crop']
-
- for obj_name in self.skeleton2d['obj_names']:
-
- # Detect body parts in wrong view
- for label in self.skeleton2d[obj_name]['label_names']:
- is_nan, new_camns = {}, {}
- sum_in_out = np.zeros(np.size(self.skeleton2d[obj_name]['frames']))
- for camn in range(1, self.nb_cam + 1):
- is_nan[camn] = np.isnan(self.skeleton2d[obj_name][label][camn]['x'])
-
- # in_out will be 0 if the camera view is correct, otherwise an int (-1, 1, 2, etc)
- in_out_x = np.ceil((self.skeleton2d[obj_name][label][camn]['x'] - (camn-1)*width_crop)/width_crop) -1
- in_out_y = np.ceil((self.skeleton2d[obj_name][label][camn]['y'])/height_crop) -1
-
- new_camns[camn] = np.ones(np.size(in_out_x)) *camn
- for i, frame in enumerate(self.skeleton2d[obj_name]['frames']): # assumes that stitching was done along x axis
- if (in_out_x[i] + camn) < 1 or (in_out_x[i] + camn) > self.nb_cam: new_camns[camn][i] = np.nan
- elif in_out_y[i] < 0 or in_out_y[i] > 0: new_camns[camn][i] = np.nan
- else: new_camns[camn][i] = in_out_x[i] + camn
-
- sum_in_out += new_camns[camn] - camn
-
- skeleton2d_cur_bodypart = copy.deepcopy(self.skeleton2d[obj_name][label])
- for camn in range(1, self.nb_cam + 1):
- skeleton2d_cur_bodypart[camn]['x'] *= np.nan
- skeleton2d_cur_bodypart[camn]['y'] *= np.nan
-
- for i, frame in enumerate(self.skeleton2d[obj_name]['frames']):
- for camn in range(1, self.nb_cam + 1):
- new_camn = new_camns[camn][i]
- if np.isnan(new_camn) or is_nan[camn][i]: continue
-
- if sum_in_out[i] == 0 or (is_nan[new_camn][i] and sum_in_out[i] != 0):
- skeleton2d_cur_bodypart[new_camn]['x'][i] = self.skeleton2d[obj_name][label][camn]['x'][i]
- skeleton2d_cur_bodypart[new_camn]['y'][i] = self.skeleton2d[obj_name][label][camn]['y'][i]
-
- elif not np.isnan(new_camns[new_camn][i]):
- skeleton2d_cur_bodypart[new_camn]['x'][i] = self.skeleton2d[obj_name][label][new_camn]['x'][i]
- skeleton2d_cur_bodypart[new_camn]['y'][i] = self.skeleton2d[obj_name][label][new_camn]['y'][i]
-
- # if self.show_plot:
- # fig, axs = plt.subplots(1, self.nb_cam)
- # for i, camn in enumerate(range(1, self.nb_cam + 1)):
- # axs[i].clear()
- # axs[i].plot(self.skeleton2d[obj_name][label][camn]['x'], self.skeleton2d[obj_name][label][camn]['y'])
- # axs[i].plot(skeleton2d_cur_bodypart[camn]['x'], skeleton2d_cur_bodypart[camn]['y'])
- # axs[i].set_ylabel('y pos (m) - ' + label)
- # axs[i].set_xlabel('x pos (m)')
- # plt.show()
-
- self.skeleton2d[obj_name][label] = skeleton2d_cur_bodypart
-
- def unscramble_sides2d_and_recon3d_dlc(self):
- # do not work as planed
-
- reconstructor = Reconstructor3D(self.nb_cam, self.img_size)
- reconstructor.dlt_coefs = self.dlt_coefs
-
- # Unscamble case where side is wrong on only a minority of cam views
- self.skeleton3d = {'obj_names': self.skeleton2d['obj_names'], 'model_name': self.skeleton2d['model_name']}
- for obj_name in self.skeleton2d['obj_names']:
- label_names_right = [s for s in self.skeleton2d[obj_name]['label_names'] if 'right' in s]
- other_label_names = [s for s in self.skeleton2d[obj_name]['label_names'] if
- 'right' not in s or 'left' not in s]
-
- self.skeleton3d[obj_name] = {'frames': self.skeleton2d[obj_name]['frames'],
- 'label_names': self.skeleton2d[obj_name]['label_names']}
- for label in self.skeleton3d[obj_name]['label_names']:
- self.skeleton3d[obj_name][label] = {'x': np.nan * np.ones(np.size(self.skeleton2d[obj_name]['frames'])),
- 'y': np.nan * np.ones(np.size(self.skeleton2d[obj_name]['frames'])),
- 'z': np.nan * np.ones(np.size(self.skeleton2d[obj_name]['frames']))}
-
- # Compute angles btw the right and left wing/bodyparts
- angles, angles_out, mean_angles = {}, {}, {}
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- angles[camn], angles_out[camn] = {}, {}
- mean_angles[camn] = np.zeros(np.size(self.skeleton2d[obj_name][label_names_right[0]][camn]['x']))
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
- label = label_right.replace('right', '')
-
- x_right = self.skeleton2d[obj_name][label_right][camn]['x']
- y_right = self.skeleton2d[obj_name][label_right][camn]['y']
- x_left = self.skeleton2d[obj_name][label_left][camn]['x']
- y_left = self.skeleton2d[obj_name][label_left][camn]['y']
-
- angles[camn][label] = np.arctan2(y_right - y_left, x_right - x_left)
- #angles[camn][label] = np.arctan((y_right - y_left) / (x_right - x_left))
-
- angles[camn][label] = abs(angles[camn][label]) # doesn't not work when pi jump btw pi/2 and -pi/2
-
- # angles[camn][label] = np.fmod(angles[camn][label], 2 * np.pi)
- # angles[camn][label][~np.isnan(angles[camn][label])] = \
- # np.unwrap(angles[camn][label][~np.isnan(angles[camn][label])], discont=2 * np.pi)
- # # angles[camn][label][~np.isnan(angles[camn][label])] = \
- # # np.unwrap(angles[camn][label][~np.isnan(angles[camn][label])] * 2, discont=2 * np.pi * 0.8) / 2
-
- is_nan, x = np.isnan(angles[camn][label]), lambda z: z.nonzero()[0]
- if sum(is_nan) > 0: # interpolation needed for filter_high_freq
- angles[camn][label][is_nan] = np.interp(x(is_nan), x(~is_nan), angles[camn][label][~is_nan])
-
- w = self.cutoff_frequency / (self.framerate / 2) # Normalize the frequency
- b, a = signal.butter(2, w, 'low')
-
- mean_angles[camn] += signal.filtfilt(b, a, angles[camn][label])
-
- mean_angles[camn] = mean_angles[camn] / len(label_names_right)
-
- min_angle = 0
- for label_right in label_names_right:
- label = label_right.replace('right', '')
- in_out = np.where(abs(angles[camn][label] - mean_angles[camn]) > np.pi / 2)
- angles_out[camn][label] = [s if i in in_out[0] else np.nan for i, s in enumerate(angles[camn][label])]
- min_angle = np.min([np.min(angles[camn][label]), min_angle])
-
- if self.show_plot:
- fig1, axs1 = plt.subplots(self.nb_cam, 1)
- fig1.suptitle('Angle of the line btw right and left wing part - Before unscrambling')
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- axs1[i].clear()
- for label_right in label_names_right:
- label = label_right.replace('right', '')
- frames = list(range(0, len(angles_out[camn][label])))
-
- axs1[i].scatter(frames, angles[camn][label], label=label, marker='.')
- axs1[i].scatter(frames, angles_out[camn][label], c='r', label='out', marker='*')
-
- axs1[i].plot(mean_angles[camn], 'k', label='mean')
- axs1[i].set_ylabel('angle (rad)')
- axs1[i].set_xlabel('frames - ' + label_right)
-
- for i, frame in enumerate(self.skeleton2d[obj_name]['frames']):
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
- label = label_right.replace('right', '')
-
- #if all([np.isnan(angles_out[camn][label][i]) for camn in range(1, self.nb_cam + 1)]): continue
-
- xs_coord, ys_coord, frames_coord, likelihood_coords = [], [], [], []
- for camn in range(1, self.nb_cam + 1):
- if not np.isnan(angles_out[camn][label][i]): # Switch side
- self.skeleton2d[obj_name][label_right][camn]['x'][i], self.skeleton2d[obj_name][label_left][camn]['x'][i] = \
- self.skeleton2d[obj_name][label_left][camn]['x'][i], self.skeleton2d[obj_name][label_right][camn]['x'][i]
- self.skeleton2d[obj_name][label_right][camn]['y'][i], self.skeleton2d[obj_name][label_left][camn]['y'][i] = \
- self.skeleton2d[obj_name][label_left][camn]['y'][i], self.skeleton2d[obj_name][label_right][camn]['y'][i]
- self.skeleton2d[obj_name][label_right][camn]['likelihood'][i], self.skeleton2d[obj_name][label_left][camn]['likelihood'][i] = \
- self.skeleton2d[obj_name][label_left][camn]['likelihood'][i], self.skeleton2d[obj_name][label_right][camn]['likelihood'][i]
-
- x_right = self.skeleton2d[obj_name][label_right][camn]['x'][i].copy()
- y_right = self.skeleton2d[obj_name][label_right][camn]['y'][i].copy()
- x_left = self.skeleton2d[obj_name][label_left][camn]['x'][i].copy()
- y_left = self.skeleton2d[obj_name][label_left][camn]['y'][i].copy()
-
- likelihood_right = self.skeleton2d[obj_name][label_right][camn]['likelihood'][i]
- likelihood_left = self.skeleton2d[obj_name][label_left][camn]['likelihood'][i]
-
- likelihood_coords.append([likelihood_right, likelihood_left])
- xs_coord.append([x_right, x_left])
- ys_coord.append([y_right, y_left])
- frames_coord.append([frame, frame])
-
- xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
- reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False, sort_by_rmse=False)
-
- # if len(xs_pose) == 0:
- # xs_pose, ys_pose, zs_pose = [[np.nan, np.nan]], [[np.nan, np.nan]], [[np.nan, np.nan]]
- # indexes = [[np.array([0, 0, 0]), np.array([1, 1, 1])]]
- #
- # else:
- # rmses
- # # order rmse
- #
- # cpt = 0
- # while len(xs_pose[0]) > cpt:
- # # Remove duplicate usage of points and give priority to first rows (with lowest rmse)
- # # should have max two rows in the end (one per side)
- #
- # comparison_indexes = np.array([False] * self.nb_cam) # False if different from previous rows
- # for j in range(1, cpt + 1):
- # comparison_indexes += np.array(
- # [indexes[0][cpt][k] == indexes[0][cpt - j][k] for k in range(0, self.nb_cam)])
- #
- # # Remove coordinates because it use at least one 2d points that was used already
- # if comparison_indexes.any():
- # xs_pose[0] = [s for k, s in enumerate(xs_pose[0]) if k != cpt]
- # ys_pose[0] = [s for k, s in enumerate(ys_pose[0]) if k != cpt]
- # indexes[0] = [s for k, s in enumerate(indexes[0]) if k != cpt]
- # else:
- # cpt += 1
- #
- # if len(xs_pose[0]) == 1:
- # xs_pose[0].append(np.nan), ys_pose[0].append(np.nan), zs_pose[0].append(np.nan)
- # indexes[0].append([abs(index - 1) for index in indexes[0]][0])
- #
- # # if np.nansum(indexes[0][0]) > np.nansum(indexes[0][1]) \
- # # and indexes[0][1][1] == 0: # use top cam (#2) as reference for identifying side
- # if indexes[0][1][1] == 0: # use top cam (#2) as reference for identifying side
- # indexes[0][0], indexes[0][1] = indexes[0][1], indexes[0][0]
- # xs_pose[0].reverse(), ys_pose[0].reverse(), zs_pose[0].reverse()
-
- if len(xs_pose) == 0:
- xs_pose, ys_pose, zs_pose = [[np.nan, np.nan]], [[np.nan, np.nan]], [[np.nan, np.nan]]
- indexes = [[np.array([0, 0, 0]), np.array([1, 1, 1])]]
-
- elif len(xs_pose[0]) == 1:
- xs_pose[0].append(np.nan), ys_pose[0].append(np.nan), zs_pose[0].append(np.nan)
- indexes[0].append([abs(index - 1) for index in indexes[0]][0])
- #
- # elif len(xs_pose[0]) > 2:
- # indexes_top_right = np.array([s for s in indexes[0] if s[self.top_camn - 1] == 0.0]).astype(int)
- # indexes_top_left = np.array([s for s in indexes[0] if s[self.top_camn - 1] == 1.0]).astype(int)
- # rmses_top_right = np.array([rmses[0][i] for i, s in enumerate(indexes[0]) if s[self.top_camn - 1] == 0.0])
- # rmses_top_left = np.array([rmses[0][i] for i, s in enumerate(indexes[0]) if s[self.top_camn - 1] == 1.0])
- #
- # indexes = [[np.array([0, 0, 0]), np.array([1, 1, 1])]]
- # xs_pose_new, ys_pose_new, zs_pose_new = [[np.nan, np.nan]], [[np.nan, np.nan]], [[np.nan, np.nan]]
- # if len(rmses_top_right) > 1 and len(rmses_top_left) > 1:
- # combi = list(itertools.product(list(range(0, len(rmses_top_right))), list(range(0, len(rmses_top_left)))))
- #
- # new_combi = []
- # for j in range(0, len(combi)): # Keep only the combinations that are compatible
- # if (indexes_top_right[combi[j][0]] + indexes_top_left[combi[j][1]] == [1, 1, 1]).all():
- # new_combi.append(combi[j])
- #
- # if not len(new_combi) == 0:
- # rmse_combi = [np.sqrt(rmses_top_right[s[0]] ** 2 + rmses_top_left[s[1]] ** 2) for s in new_combi]
- # best_combi = new_combi[np.argsort(rmse_combi)[0]]
- #
- # indexes = [[indexes_top_right[best_combi[0]], indexes_top_left[best_combi[1]]]]
- #
- # else:
- # in_sort_right = np.argsort(rmses_top_right)
- # in_sort_left = np.argsort(rmses_top_left)
- #
- # if rmses_top_right[in_sort_right[0]] < rmses_top_left[in_sort_left[0]]:
- # indexes = [[indexes_top_right[in_sort_right[0]], abs(indexes_top_right[in_sort_right[0]] -1)]]
- # else:
- # indexes = [[abs(indexes_top_left[in_sort_left[0]] -1), indexes_top_left[in_sort_left[0]]]]
- #
- # xs_pose_new[0][0], ys_pose_new[0][0], zs_pose_new[0][0] = \
- # xs_pose[0][indexes[0][0][0]], ys_pose[0][indexes[0][0][1]], zs_pose[0][indexes[0][0][2]]
- # xs_pose_new[0][1], ys_pose_new[0][1], zs_pose_new[0][1] = \
- # xs_pose[0][indexes[0][1][0]], ys_pose[0][indexes[0][1][1]], zs_pose[0][indexes[0][1][2]]
- #
- # elif not len(rmses_top_right) == 0 and len(rmses_top_left) == 0:
- # in_sort_right = np.argsort(rmses_top_right)
- # in_sort_left = np.argsort(rmses_top_left)
- #
- # if len(rmses_top_right) == 0:
- # indexes[0][0] = abs(indexes_top_left[in_sort_left[0]] - 1)
- # else:
- # indexes[0][0] = indexes_top_right[in_sort_right[0]]
- # xs_pose_new[0][0], ys_pose_new[0][0], zs_pose_new[0][0] = \
- # xs_pose[0][indexes[0][0][0]], ys_pose[0][indexes[0][0][1]], zs_pose[0][indexes[0][0][2]]
- #
- # if len(rmses_top_left) == 0:
- # indexes[0][1] = abs(indexes_top_right[in_sort_right[0]] - 1)
- # else:
- # indexes[0][1] = indexes_top_left[in_sort_left[0]]
- # xs_pose_new[0][1], ys_pose_new[0][1], zs_pose_new[0][1] = \
- # xs_pose[0][indexes[0][1][0]], ys_pose[0][indexes[0][1][1]], zs_pose[0][indexes[0][1][2]]
- #
- # xs_pose, ys_pose, zs_pose = xs_pose_new, ys_pose_new, zs_pose_new
- #
- # if indexes[0][1][self.top_camn-1] == 0: # use top cam as reference for identifying side
- # indexes[0][0], indexes[0][1] = indexes[0][1], indexes[0][0]
- # xs_pose[0].reverse(), ys_pose[0].reverse(), zs_pose[0].reverse()
- #
- # if (indexes[0] != np.array([0, 0, 0])).any():
- # for j, camn in enumerate(range(1, self.nb_cam + 1)):
- # if not np.isnan(indexes[0][0][j]) and not np.isnan(indexes[0][1][j]):
- # in_right = int(indexes[0][0][j])
- # self.skeleton2d[obj_name][label_right][camn]['x'][i] = xs_coord[j][in_right]
- # self.skeleton2d[obj_name][label_right][camn]['y'][i] = ys_coord[j][in_right]
- # self.skeleton2d[obj_name][label_right][camn]['likelihood'][i] = likelihood_coords[j][in_right]
- #
- # in_left = int(indexes[0][1][j])
- # self.skeleton2d[obj_name][label_left][camn]['x'][i] = xs_coord[j][in_left]
- # self.skeleton2d[obj_name][label_left][camn]['y'][i] = ys_coord[j][in_left]
- # self.skeleton2d[obj_name][label_left][camn]['likelihood'][i] = likelihood_coords[j][in_left]
-
- self.skeleton3d[obj_name][label_right]['x'][i], self.skeleton3d[obj_name][label_right]['y'][i], \
- self.skeleton3d[obj_name][label_right]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
-
- self.skeleton3d[obj_name][label_left]['x'][i], self.skeleton3d[obj_name][label_left]['y'][i], \
- self.skeleton3d[obj_name][label_left]['z'][i] = xs_pose[0][1], ys_pose[0][1], zs_pose[0][1]
-
- for label in other_label_names:
- xs_coord, ys_coord, frames_coord = [], [], []
- for camn in range(1, self.nb_cam + 1):
- xs_coord.append([self.skeleton2d[obj_name][label][camn]['x'][i].copy()])
- ys_coord.append([self.skeleton2d[obj_name][label][camn]['y'][i].copy()])
- frames_coord.append([frame])
-
- xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
- reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
-
- if len(xs_pose) == 0:
- self.skeleton3d[obj_name][label]['x'][i], self.skeleton3d[obj_name][label]['y'][i], \
- self.skeleton3d[obj_name][label]['z'][i] = np.nan, np.nan, np.nan
- else:
- self.skeleton3d[obj_name][label]['x'][i], self.skeleton3d[obj_name][label]['y'][i], \
- self.skeleton3d[obj_name][label]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
-
- if self.show_plot:
- fig2, axs2 = plt.subplots(self.nb_cam, 1)
- fig2.suptitle('Angle of the line btw right and left wing part - After unscrambling')
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- axs2[i].clear()
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
- x_right = self.skeleton2d[obj_name][label_right][camn]['x']
- y_right = self.skeleton2d[obj_name][label_right][camn]['y']
- x_left = self.skeleton2d[obj_name][label_left][camn]['x']
- y_left = self.skeleton2d[obj_name][label_left][camn]['y']
-
- angles = np.arctan2((y_right - y_left), (x_right - x_left))
- #angles = np.arctan((y_right - y_left) / (x_right - x_left))
- #angles[~np.isnan(angles)] = np.unwrap(angles[~np.isnan(angles)] * 2, discont=2 * np.pi * 0.8) / 2
- angles = np.mod(angles, 2*np.pi)
-
- axs2[i].plot(angles, label=label_right.replace('right', ''))
- axs2[i].set_ylabel('angle (rad)')
- axs2[i].set_xlabel('frames - ' + label_right)
- plt.legend
- plt.show()
-
- def unscramble_sides3d_dlc(self, body_params):
- # Unscramble case where side is wrong on all cam views
- for obj_name in self.skeleton3d['obj_names']:
- label_names_right = [s for s in self.skeleton3d[obj_name]['label_names'] if 'right' in s]
-
- if (self.skeleton2d[obj_name]['frames'] != self.skeleton3d[obj_name]['frames']).any():
- raise ValueError('frames should be the same for skeleton2d and skeleton3d')
-
- if self.show_plot:
- fig1, axs1 = plt.subplots(self.nb_cam, 1)
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- axs1[i].clear()
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
- x_right = self.skeleton2d[obj_name][label_right][camn]['x']
- y_right = self.skeleton2d[obj_name][label_right][camn]['y']
- x_left = self.skeleton2d[obj_name][label_left][camn]['x']
- y_left = self.skeleton2d[obj_name][label_left][camn]['y']
-
- axs1[i].plot(np.arctan2(y_right - y_left, x_right - x_left)) # angle of the line btw the two sides
- axs1[i].set_ylabel('angle btw right and left (rad)')
- axs1[i].set_xlabel('frames - ' + label_right)
-
- for i, frame in enumerate(self.skeleton3d[obj_name]['frames']):
- body_params_frame = {}
- for label in body_params[obj_name].keys():
- body_params_frame[label] = body_params[obj_name][label][i].copy()
-
- body_skeleton3d = copy.deepcopy(self.body_skeleton3d_init)
- body_skeleton3d = build_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton3d = rotate_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton3d = translate_body_skeleton3d(body_skeleton3d, body_params_frame)
- # body_skeleton2d = reproject_skeleton3d_to2d(body_skeleton3d, self.dlt_coefs)
-
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
-
- x_right = self.skeleton3d[obj_name][label_right]['x'][i].copy()
- y_right = self.skeleton3d[obj_name][label_right]['y'][i].copy()
- z_right = self.skeleton3d[obj_name][label_right]['z'][i].copy()
-
- x_left = self.skeleton3d[obj_name][label_left]['x'][i].copy()
- y_left = self.skeleton3d[obj_name][label_left]['y'][i].copy()
- z_left = self.skeleton3d[obj_name][label_left]['z'][i].copy()
-
- xyz_right = [x_right, y_right, z_right]
- xyz_left = [x_left, y_left, z_left]
-
- dist_right2right_hinge = np.linalg.norm(body_skeleton3d['right_wing_hinge'] - xyz_right)
- dist_left2right_hinge = np.linalg.norm(body_skeleton3d['right_wing_hinge'] - xyz_left)
- dist_right2left_hinge = np.linalg.norm(body_skeleton3d['left_wing_hinge'] - xyz_right)
- dist_left2left_hinge = np.linalg.norm(body_skeleton3d['left_wing_hinge'] - xyz_left)
-
- if dist_right2right_hinge > dist_right2left_hinge and dist_left2left_hinge > dist_left2right_hinge:
- x_right, y_right, z_right, x_left, y_left, z_left = x_left, y_left, z_left, x_right, y_right, z_right
-
- self.skeleton3d[obj_name][label_right]['x'][i], self.skeleton3d[obj_name][label_right]['y'][i], \
- self.skeleton3d[obj_name][label_right]['z'][i] = x_right, y_right, z_right
-
- self.skeleton3d[obj_name][label_left]['x'][i], self.skeleton3d[obj_name][label_left]['y'][i], \
- self.skeleton3d[obj_name][label_left]['z'][i] = x_left, y_left, z_left
-
- for camn in range(1, self.nb_cam + 1):
- self.skeleton2d[obj_name][label_right][camn]['x'][i], \
- self.skeleton2d[obj_name][label_right][camn]['y'][i], \
- self.skeleton2d[obj_name][label_left][camn]['x'][i], \
- self.skeleton2d[obj_name][label_left][camn]['y'][i] = \
- self.skeleton2d[obj_name][label_left][camn]['x'][i], \
- self.skeleton2d[obj_name][label_left][camn]['y'][i], \
- self.skeleton2d[obj_name][label_right][camn]['x'][i], \
- self.skeleton2d[obj_name][label_right][camn]['y'][i]
-
- if self.show_plot:
- fig2, axs2 = plt.subplots(self.nb_cam, 1)
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- axs2[i].clear()
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
- x_right = self.skeleton2d[obj_name][label_right][camn]['x']
- y_right = self.skeleton2d[obj_name][label_right][camn]['y']
- x_left = self.skeleton2d[obj_name][label_left][camn]['x']
- y_left = self.skeleton2d[obj_name][label_left][camn]['y']
-
- axs2[i].plot(
- np.arctan2(y_right - y_left, x_right - x_left)) # angle of the line btw the two sides
- axs2[i].set_ylabel('angle btw right and left (rad)')
- axs2[i].set_xlabel('frames - ' + label_right)
-
- plt.show()
-
- def reverse_processes_2d_points(self, process_dict):
- [width_img, height_img] = self.img_size
- for camn in range(1, self.nb_cam + 1):
- process_names = [process_dict[camn][process_num]['fn'] for process_num in process_dict[camn].keys()]
-
- for obj_name in self.skeleton2d['obj_names']:
- if 'stitch' in process_names or 'rotate' in process_names:
- if 'crop' in process_names:
- crop_num = process_names.index('crop')
- crop_num = crop_num[-1] if isinstance(crop_num, list) else crop_num
- width_crop = process_dict[camn][crop_num]['kwargs']['width_crop']
- height_crop = process_dict[camn][crop_num]['kwargs']['height_crop']
-
- else:
- width_crop, height_crop = width_img, height_img
-
- if 'crop' in process_names:
- crop_path = self.get_path_process('crop')[process_dict[camn][0]['camn']]
- csv_path = glob.glob(os.path.join(crop_path, process_dict[camn][0]['rec_names'][camn],
- '*' + obj_name + '-2d_points.csv'))
-
- pts_csv = np.genfromtxt(csv_path[0], delimiter=',', skip_header=0, names=True)
-
- temp_set = set(self.skeleton2d[obj_name]['frames'])
- in_frame = [i for i, val in enumerate(pts_csv['frame']) if val in temp_set]
-
- width_crop = process_dict[camn][crop_num]['kwargs']['width_crop']
- height_crop = process_dict[camn][crop_num]['kwargs']['height_crop']
-
- x_crop = pts_csv['x_px'][in_frame] - width_crop / 2
- y_crop = pts_csv['y_px'][in_frame] - height_crop / 2
-
- for i in range(0, len(in_frame)):
- x_crop[i] = np.max([np.min([x_crop[i], width_img - width_crop]), 0])
- y_crop[i] = np.max([np.min([y_crop[i], height_img - height_crop]), 0])
-
- for label in self.skeleton2d[obj_name]['label_names']:
- if 'stitch' in process_names:
- self.skeleton2d[obj_name][label][camn]['x'] += - width_crop * (camn - 1)
-
- if 'rotate' in process_names and \
- camn in process_dict[camn][process_names.index('rotate')]['kwargs']['camns']:
- x = self.skeleton2d[obj_name][label][camn]['x'].copy() - width_crop / 2
- y = self.skeleton2d[obj_name][label][camn]['y'].copy() - height_crop / 2
- teta = np.deg2rad(process_dict[camn][process_names.index('rotate')]['kwargs']['degrees'])
-
- self.skeleton2d[obj_name][label][camn]['x'] = x * np.cos(teta) - y * np.sin(teta) + width_crop / 2
- self.skeleton2d[obj_name][label][camn]['y'] = x * np.sin(teta) + y * np.cos(teta) + height_crop / 2
-
- if 'crop' in process_names:
- self.skeleton2d[obj_name][label][camn]['x'] += x_crop
- self.skeleton2d[obj_name][label][camn]['y'] += y_crop
-
- if self.show_plot:
- raw_rec_path = os.path.join(process_dict[camn][0]['paths'][process_dict[camn][0]['camn']],
- process_dict[camn][0]['rec_names'][process_dict[camn][0]['camn']])
- all_files = sorted(os.listdir(raw_rec_path), key=lambda x: x[20:20 + self.nb_leading_zero])
- all_images = [s for s in all_files if '.{0}'.format(self.image_format) in s]
-
- fig = plt.figure()
- for i, image in enumerate(all_images):
- if i % 500: continue # plot every 100 frames
- ax = fig.add_subplot(111)
- plt.clf()
- plt.imshow(Image.open(os.path.join(raw_rec_path, image)))
- # ax.scatter(x_crop + width_crop/2, y_crop + height_crop/2, marker='o')
- for label in self.skeleton2d[obj_name]['label_names']:
- # if 'wing' not in label: continue
- if self.skeleton2d[obj_name][label][camn]['likelihood'][i] > self.threshold_likelihood:
- ax.scatter(self.skeleton2d[obj_name][label][camn]['x'][i],
- self.skeleton2d[obj_name][label][camn]['y'][i], 1)
- ax.set_xlabel('x (m)')
- ax.set_ylabel('y (m)')
- plt.pause(0.1)
- # plt.show()
-
- def test_unwrap(self):
- reconstructor = Reconstructor3D(self.nb_cam, self.img_size)
- reconstructor.dlt_coefs = self.dlt_coefs
-
- # Unscamble case where side is wrong on only a minority of cam views
- self.skeleton3d = {'obj_names': self.skeleton2d['obj_names'], 'model_name': self.skeleton2d['model_name']}
- for obj_name in self.skeleton3d['obj_names']:
- label_names_right = [s for s in self.skeleton2d[obj_name]['label_names'] if 'right' in s]
-
- self.skeleton3d[obj_name] = {'frames': self.skeleton2d[obj_name]['frames'],
- 'label_names': self.skeleton2d[obj_name]['label_names']}
- for label in self.skeleton3d[obj_name]['label_names']:
- self.skeleton3d[obj_name][label] = {'x': np.nan * np.ones(np.size(self.skeleton3d[obj_name]['frames'])),
- 'y': np.nan * np.ones(np.size(self.skeleton3d[obj_name]['frames'])),
- 'z': np.nan * np.ones(np.size(self.skeleton3d[obj_name]['frames']))}
-
- for i, frame in enumerate(self.skeleton3d[obj_name]['frames']):
- for label in self.skeleton3d[obj_name]['label_names']:
- xs_coord, ys_coord, frames_coord = [], [], []
- for camn in range(1, self.nb_cam + 1):
- xs_coord.append([self.skeleton2d[obj_name][label][camn]['x'][i].copy()])
- ys_coord.append([self.skeleton2d[obj_name][label][camn]['y'][i].copy()])
- frames_coord.append([frame])
-
- xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
- reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
-
- if len(xs_pose) == 0:
- self.skeleton3d[obj_name][label]['x'][i], self.skeleton3d[obj_name][label]['y'][i], \
- self.skeleton3d[obj_name][label]['z'][i] = np.nan, np.nan, np.nan
- else:
- self.skeleton3d[obj_name][label]['x'][i], self.skeleton3d[obj_name][label]['y'][i], \
- self.skeleton3d[obj_name][label]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
-
- fig1, axs1 = plt.subplots(self.nb_cam, 1)
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- axs1[i].clear()
- for label_right in label_names_right:
- label_left = label_right.replace('right', 'left')
- x_right = self.skeleton2d[obj_name][label_right][camn]['x']
- y_right = self.skeleton2d[obj_name][label_right][camn]['y']
- x_left = self.skeleton2d[obj_name][label_left][camn]['x']
- y_left = self.skeleton2d[obj_name][label_left][camn]['y']
-
- angle = np.arctan((y_right - y_left) / (x_right - x_left))
- angle[~np.isnan(angle)] = np.unwrap(angle[~np.isnan(angle)] * 2, discont=2 * np.pi * 0.8) / 2
-
- angle2 = np.arctan2((y_right - y_left), (x_right - x_left))
- median_diff_angle = np.nanmedian(angle2 - angle)
- if abs(abs(median_diff_angle) - np.pi) < 0.2:
- if median_diff_angle > 0:
- angle2 -= np.pi
- elif median_diff_angle < 0:
- angle2 += np.pi
-
- in_diff = np.array([False] * len(angle))
- is_not_nan = ~np.isnan(angle) & ~np.isnan(angle2)
- in_diff[is_not_nan] = abs(angle2[is_not_nan] - angle[is_not_nan]) > np.pi / 2
-
- axs1[i].plot(self.skeleton2d[obj_name]['frames'], angle) # angle of the line btw the two sides
- axs1[i].plot(np.array(self.skeleton2d[obj_name]['frames'])[in_diff], angle2[in_diff], marker='*',
- linestyle='None')
-
- axs1[i].set_ylabel('angle btw right and left (rad)')
- axs1[i].set_xlabel('frames - ' + label_right)
-
- plt.show()
-
- def recon3d_dlc(self):
- reconstructor = Reconstructor3D(self.nb_cam, self.img_size)
- reconstructor.dlt_coefs = self.dlt_coefs
-
- # Unscamble case where side is wrong on only a minority of cam views
- self.skeleton3d = {'obj_names': self.skeleton2d['obj_names'], 'model_name': self.skeleton2d['model_name']}
- for obj_name in self.skeleton3d['obj_names']:
- # label_names_right = [s for s in self.skeleton2d[obj_name]['label_names'] if 'right' in s]
- # other_label_names = [s for s in self.skeleton2d[obj_name]['label_names'] if
- # 'right' not in s or 'left' not in s]
-
- self.skeleton3d[obj_name] = {'frames': self.skeleton2d[obj_name]['frames'],
- 'label_names': self.skeleton2d[obj_name]['label_names']}
- for label in self.skeleton3d[obj_name]['label_names']:
- self.skeleton3d[obj_name][label] = {'x': np.nan * np.ones(np.size(self.skeleton2d[obj_name]['frames'])),
- 'y': np.nan * np.ones(np.size(self.skeleton2d[obj_name]['frames'])),
- 'z': np.nan * np.ones(np.size(self.skeleton2d[obj_name]['frames']))}
-
- for i, frame in enumerate(self.skeleton2d[obj_name]['frames']):
- for label in self.skeleton3d[obj_name]['label_names']:
- xs_coord, ys_coord, frames_coord = [], [], []
- for camn in range(1, self.nb_cam + 1):
- xs_coord.append([self.skeleton2d[obj_name][label][camn]['x'][i].copy()])
- ys_coord.append([self.skeleton2d[obj_name][label][camn]['y'][i].copy()])
- frames_coord.append([frame])
-
- xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
- reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
-
- if len(xs_pose) == 0:
- self.skeleton3d[obj_name][label]['x'][i], self.skeleton3d[obj_name][label]['y'][i], \
- self.skeleton3d[obj_name][label]['z'][i] = np.nan, np.nan, np.nan
- else:
- self.skeleton3d[obj_name][label]['x'][i], self.skeleton3d[obj_name][label]['y'][i], \
- self.skeleton3d[obj_name][label]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
-
- def save2d_dlc(self, save_paths):
- for obj_name in self.skeleton2d['obj_names']:
- for label in self.skeleton2d[obj_name]['label_names']:
- for camn in range(1, self.nb_cam + 1):
- if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
- np.savetxt(os.path.join(save_paths[camn], obj_name + '-' + label + '-2d_points.csv'),
- np.c_[self.skeleton2d[obj_name]['frames'], self.skeleton2d[obj_name][label][camn]['x'],
- self.skeleton2d[obj_name][label][camn]['y'],
- self.skeleton2d[obj_name][label][camn]['likelihood']],
- delimiter=',', header='frame,x_px,y_px,likelihood')
-
- # # To save all in same .csv file
- # for camn in range(1, self.nb_cam + 1):
- # if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
- # header = header + ', x{0}_px, y{0}_px, likelihood{0}'.format(camn)
- # array_csv = np.concatenate((array_csv, np.c_[self.skeleton2d[obj_name][label][camn]['x'],
- # self.skeleton2d[obj_name][label][camn]['y'],
- # self.skeleton2d[obj_name][label][camn]['likelihood']), axis=1)
- #
- # np.savetxt(os.path.join(save_paths[self.main_camn], obj_name + '-' + label + '-2d_points.csv'),
- # array_csv, delimiter=',', header=header)
-
- def save3d_dlc(self, save_paths):
- for obj_name in self.skeleton3d['obj_names']:
- for label in self.skeleton3d[obj_name]['label_names']:
- for camn in range(1, self.nb_cam + 1):
- if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
- np.savetxt(os.path.join(save_paths[camn], obj_name + '-' + label + '-3d_points.csv'),
- np.c_[self.skeleton3d[obj_name]['frames'], self.skeleton3d[obj_name][label]['x'],
- self.skeleton3d[obj_name][label]['y'], self.skeleton3d[obj_name][label]['z']],
- delimiter=',', header='frame,x,y,z')
-
- def estimate_body_wings_params(self, interpolate_nans=False, filter_high_freq=False):
- body_params_labels = self.body_params_init.keys()
- wing_params_labels = self.wings_params_init['right'].keys()
-
- if len(self.skeleton3d['obj_names']) == 0:
- raise ValueError('obj_names is empty')
-
- self.body_params, self.wings_params, self.mean_body_params, self.mean_wings_params = {}, {}, {}, {}
- for obj_name in self.skeleton3d['obj_names']:
- frames = self.skeleton3d[obj_name]['frames']
-
- self.body_params[obj_name], self.wings_params[obj_name] = {}, {}
- self.mean_body_params[obj_name], self.mean_wings_params[obj_name] = {}, {}
- for label in body_params_labels:
- self.body_params[obj_name][label] = np.nan * np.ones(frames.shape)
-
- for side in ['right', 'left']:
- self.wings_params[obj_name][side] = {}
- for label in wing_params_labels:
- self.wings_params[obj_name][side][label] = np.nan * np.ones(frames.shape)
-
- for i, frame in enumerate(frames):
- #for label in self.skeleton3d[obj_name]['label_names']:
-
- body_skeleton3d, wings_skeleton3d = {}, {'right': {}, 'left': {}}
- for body_label in self.body_labels:
- body_skeleton3d[body_label] = np.array([self.skeleton3d[obj_name][body_label]['x'][i],
- self.skeleton3d[obj_name][body_label]['y'][i],
- self.skeleton3d[obj_name][body_label]['z'][i]])
-
- if self.keep_init_aspect_ratio: init_aspect_ratio = {}
-
- for side in ['right', 'left']:
- for wing_label in self.wing_labels:
- label = side + '_wing_' + wing_label
- wings_skeleton3d[side][wing_label] = \
- np.array([self.skeleton3d[obj_name][label]['x'][i],
- self.skeleton3d[obj_name][label]['y'][i],
- self.skeleton3d[obj_name][label]['z'][i]])
-
- if self.keep_init_aspect_ratio:
- span_v = self.wings_skeleton3d_init[side]['tip'] - self.wings_skeleton3d_init[side]['hinge']
- chord_v = self.wings_skeleton3d_init[side]['leading_edge_q2'] - self.wings_skeleton3d_init[side]['trailing_edge_q2']
- init_aspect_ratio[side] = np.round(length_from_vector(span_v), 9) / np.round(length_from_vector(chord_v), 9)
- init_aspect_ratio[side] = init_aspect_ratio[side] * np.ones(np.shape(self.wings_params[obj_name][side]['span']))
-
- # for leg_name in leg_names:
- # for leg_label in leg_labels:
- # label = side + '_leg_' + leg_name + '_' + leg_label
- # body_skeleton3d[label] = np.array([self.skeleton3d[obj_name][label]['x'][i],
- # self.skeleton3d[obj_name][label]['y'][i],
- # self.skeleton3d[obj_name][label]['z'][i]])
-
- body_params = estimate_body_parameters_from_skeleton3d(body_skeleton3d, self.body_params_init)
- for label in body_params_labels:
- self.body_params[obj_name][label][i] = body_params[label]
-
- for side in ['right', 'left']:
- wing_param = estimate_wing_parameters_from_skeleton3d(wings_skeleton3d[side], body_params,
- self.wings_params_init[side], side)
-
- for label in wing_params_labels:
- self.wings_params[obj_name][side][label][i] = wing_param[label]
-
- if self.keep_init_aspect_ratio:
- for side in ['right', 'left']:
- self.wings_params[obj_name][side]['aspect_ratio'] = init_aspect_ratio[side]
- self.wings_params[obj_name][side]['chord'] = self.wings_params[obj_name][side]['span']/init_aspect_ratio[side]
-
- # if self.show_plot:
- # # Plot body and wings params that will be kept cst
- # fig1, axs1 = plt.subplots(len(self.body_labels_to_keep_cst), 1)
- # for i, label in enumerate(self.body_labels_to_keep_cst):
- # y = self.body_params[obj_name][label]
- # axs1[i].plot(frames, y, '.', label='estim')
- # axs1[i].plot(frames, np.nanmean(y) * np.ones(frames.shape), '-', label='mean')
- # axs1[i].plot(frames, np.nanmedian(y) * np.ones(frames.shape), '-', label='median')
- # axs1[i].plot(frames, stats.mode(y)[0] * np.ones(frames.shape), '-', label='mode')
- #
- # axs1[i].set_ylabel(label)
- # axs1[i].set_xlabel('frames')
- # fig1.suptitle('Body parameters in function of time')
- # fig1.legend()
- #
- # fig2, axs2 = plt.subplots(len(self.wing_labels_to_keep_cst), 1)
- # for i, label in enumerate(self.wing_labels_to_keep_cst):
- # for side in ['right', 'left']:
- # y = self.wings_params[obj_name][side][label]
- # axs2[i].plot(frames, y, '.', label='estim_' + side)
- # axs2[i].plot(frames, np.nanmean(y) * np.ones(frames.shape), '-', label='mean_' + side)
- # axs2[i].plot(frames, np.nanmedian(y) * np.ones(frames.shape), '-', label='median_' + side)
- # axs2[i].plot(frames, stats.mode(y)[0] * np.ones(frames.shape), '-', label='mode_' + side)
- # axs2[i].set_ylabel(label)
- # axs2[i].set_xlabel('frames')
- # fig2.suptitle('Wings parameters in function of time')
- # fig2.legend()
- #
- # plt.show()
-
- # Compute median value of body and wings length (should be constant for one animal)
- for label in self.body_labels_to_keep_cst:
- self.body_params[obj_name][label] = \
- np.nanmedian(self.body_params[obj_name][label]) * np.ones(frames.shape)
- # print('>>> mean {0}: {1}'.format(label, self.body_params[obj_name][label][0]))
-
- self.body_params[obj_name] = self.filter_interp_params(self.body_params[obj_name], body_params_labels,
- interpolate_nans, filter_high_freq)
-
- for label in self.wing_labels_to_keep_cst:
- mean_over_sides = np.nanmedian([np.nanmedian(self.wings_params[obj_name][side][label]) for side in ['right', 'left']])
- for side in ['right', 'left']:
- self.wings_params[obj_name][side][label] = mean_over_sides * np.ones(frames.shape)
- # print('>>> mean {0}: {1} (right and left)'.format(label, self.wings_params[obj_name]['right'][label][0]))
-
- for side in ['right', 'left']:
- self.wings_params[obj_name][side] = \
- self.filter_interp_params(self.wings_params[obj_name][side], ['x_hinge', 'y_hinge', 'z_hinge'],
- False, filter_high_freq)
- self.wings_params[obj_name][side] = \
- self.filter_interp_params(self.wings_params[obj_name][side], wing_params_labels,
- interpolate_nans, False)
-
- # Copy all body_params in wings_params
- for side in ['right', 'left']:
- for label in body_params_labels:
- self.wings_params[obj_name][side][label] = self.body_params[obj_name][label].copy()
-
- return self.body_params, self.wings_params
-
- def estimate_wing_hinges(self, interpolate_nans=False, filter_high_freq=False):
- wing_params_labels = self.wings_params_init['right'].keys()
-
- dpt = 3
- x, y, z = symbols('x y z')
-
- if len(self.skeleton3d['obj_names']) == 0:
- raise ValueError('obj_names is empty')
-
- for obj_name in self.skeleton3d['obj_names']:
- frames = self.skeleton3d[obj_name]['frames']
-
- if self.show_plot:
- init_hinges = {}
- for side in ['right', 'left']:
- init_hinges[side] = {'x': self.wings_params[obj_name][side]['x_hinge'].copy(),
- 'y': self.wings_params[obj_name][side]['y_hinge'].copy(),
- 'z': self.wings_params[obj_name][side]['z_hinge'].copy()}
-
- for side in ['right', 'left']:
- for label in wing_params_labels:
- self.wings_params[obj_name][side][label] = np.nan * np.ones(frames.shape)
-
- for side in ['right', 'left']:
- a = np.ones(np.shape(frames)) * np.nan
- b, c, d = a.copy(), a.copy(), a.copy()
- for i, frame in enumerate(frames):
- x_coords = np.ones((len(self.wing_labels), 1)) * np.nan
- y_coords, z_coords = x_coords.copy(), x_coords.copy()
- for j, wing_label in enumerate(self.wing_labels):
- label = side + '_wing_' + wing_label
-
- x_coords[j] = self.skeleton3d[obj_name][label]['x'][i]
- y_coords[j] = self.skeleton3d[obj_name][label]['y'][i]
- z_coords[j] = self.skeleton3d[obj_name][label]['z'][i]
-
- if len(x_coords) < 3: continue
- params = minimize_perp_distance(x_coords, y_coords, z_coords)
- a[i], b[i], c[i], d[i] = params
-
- if i >= 2*dpt:
- eq = [a[i - 2*dpt] * x + b[i - 2*dpt] * y + c[i - 2*dpt] * z + d[i - 2*dpt],
- a[i - 1*dpt] * x + b[i - 1*dpt] * y + c[i - 1*dpt] * z + d[i - 1*dpt],
- a[i] * x + b[i] * y + c[i] * z + d[i]]
- sol = solve(eq, dict=True)
-
- if len(sol) > 0 and x in sol[0].keys() and y in sol[0].keys() and z in sol[0].keys():
- self.wings_params[obj_name][side]['x_hinge'][i] = sol[0][x]
- self.wings_params[obj_name][side]['y_hinge'][i] = sol[0][y]
- self.wings_params[obj_name][side]['z_hinge'][i] = sol[0][z]
-
- for side in ['right', 'left']:
- self.wings_params[obj_name][side] = \
- self.filter_interp_params(self.wings_params[obj_name][side], ['x_hinge', 'y_hinge', 'z_hinge'],
- interpolate_nans, filter_high_freq)
-
- if self.show_plot:
- fig, axs = plt.subplots(1, 3)
- for i, coord in enumerate(['x', 'y', 'z']):
- axs[i].clear()
- for side in ['right', 'left']:
- axs[i].scatter(frames, init_hinges[side][coord], marker='o', label='init_' + side)
- axs[i].scatter(frames, self.wings_params[obj_name][side][coord + '_hinge'], marker='+', label='est_' + side)
- axs[i].set_ylabel(coord + ' pos (m)')
- axs[i].set_xlabel('frames')
- plt.legend()
- plt.show()
-
- return self.wings_params
-
- def optim_fit_skeleton(self, animal_name, body_param_names, wing_param_names, res_method, opt_method, rec_paths, save_paths, angles_to_correct=[]):
- body_params_init, wings_params_init = self.estimate_body_wings_params(interpolate_nans=True, filter_high_freq=True)
- #wings_params_init = self.estimate_wing_hinges(interpolate_nans=True, filter_high_freq=True)
- self.save_skeleton_params_in_csv(body_params_init, wings_params_init, save_paths[self.main_camn])
-
- start = time.time()
- body_params, nb_visible_pts_body, rmse_body, nb_iterations_body = \
- self.optim_fit_all_hybrid_params(animal_name, body_params_init, wings_params_init, body_param_names, res_method,
- opt_method, rec_paths, interpolate_nans=False, filter_high_freq=True)
- print('>> optim hybrid done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
-
- # body_params, nb_visible_pts_body, rmse_body, nb_iterations_body = \
- # self.optim_fit_all_body_params(animal_name, body_params_init, body_param_names, res_method, opt_method, rec_paths,
- # interpolate_nans=False, filter_high_freq=True)
- # #body_params = body_params_init
- # print('>> optim body done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
-
- # self.unscramble_sides3d_dlc(body_params)
- # self.save2d_dlc(save_paths)
- # self.save3d_dlc(save_paths)
-
- start = time.time()
- wings_params, nb_visible_pts_wings, rmse_wings, nb_iterations_wings = \
- self.optim_fit_all_wings_params(animal_name, body_params, wings_params_init, wing_param_names, res_method, opt_method,
- rec_paths, interpolate_nans=False, filter_high_freq=True)
- print('>> optim wings done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
-
- #body_params, wings_params = self.correct_body_wings_params(body_params, wings_params, angles_to_correct, rec_paths)
-
- if self.show_plot:
- for obj_name in body_params.keys():
- frames = list(range(0, len(body_params[obj_name][body_param_names[0]])))
-
- # Plot body params
- fig1, axs1 = plt.subplots(len(body_param_names), 1)
- for i, label in enumerate(body_param_names):
- axs1[i].plot(frames, body_params_init[obj_name][label], '-', label='init')
- axs1[i].plot(frames, body_params[obj_name][label], '-', label='optim')
- axs1[i].set_ylabel(label)
- axs1[i].set_xlabel('frames')
- fig1.suptitle('Body parameters in function of time')
- fig1.legend()
-
- # Plot wings params
- fig2, axs2 = plt.subplots(len(wing_param_names), 1)
- for i, label in enumerate(wing_param_names):
- for side in wings_params[obj_name].keys():
- axs2[i].plot(frames, wings_params_init[obj_name][side][label], '-', label='init_' + side)
- axs2[i].plot(frames, wings_params[obj_name][side][label], '-', label='optim_' + side)
- axs2[i].set_ylabel(label)
- axs2[i].set_xlabel('frames')
- fig2.suptitle('Wings parameters in function of time')
- fig2.legend()
-
- plt.show()
-
- self.save_skeleton_params_in_csv(body_params, wings_params, save_paths[self.main_camn])
-
- self.save_obj_metric_in_csv(nb_visible_pts_body, 'nb_visible_pts-body', save_paths[self.main_camn])
- self.save_obj_metric_in_csv(nb_visible_pts_wings, 'nb_visible_pts-wings', save_paths[self.main_camn])
- self.save_obj_metric_in_csv(nb_iterations_body, 'nb_iterations-body', save_paths[self.main_camn])
- self.save_obj_metric_in_csv(nb_iterations_wings, 'nb_iterations-wings', save_paths[self.main_camn])
- self.save_obj_metric_in_csv(rmse_body, 'rmse-body', save_paths[self.main_camn])
- self.save_obj_metric_in_csv(rmse_wings, 'rmse-wings', save_paths[self.main_camn])
-
- def optim_fit_all_hybrid_params(self, animal_name, body_params_init, wings_params_init, param_names, res_method, opt_method, rec_paths,
- interpolate_nans=False, filter_high_freq=False, first_optim_wings=False):
-
- hybrid_params_init = copy.deepcopy(body_params_init)
- if '2d' in res_method: self.hybrid_skeleton2d = copy.deepcopy(self.skeleton2d)
- else: self.hybrid_skeleton3d = copy.deepcopy(self.skeleton3d)
-
- if first_optim_wings:
- start = time.time()
- # First optim on wing to estimate wing tips and hinges positions + deviation_a
- wing_param_names = ['stroke_a', 'deviation_a', 'rotation_a', 'x_hinge', 'y_hinge', 'z_hinge']
- wings_params, _, _ = \
- self.optim_fit_all_wings_params(body_params_init, wings_params_init, wing_param_names, res_method,
- opt_method, rec_paths, interpolate_nans=False, filter_high_freq=True)
-
- obj_names = self.skeleton2d['obj_names'] if '2d' in res_method else self.skeleton3d['obj_names']
- for obj_name in obj_names:
-
- frames = self.skeleton2d[obj_name]['frames'] if '2d' in res_method else self.skeleton3d[obj_name]['frames']
- for i, frame in enumerate(frames):
-
- wings_params_frame = {'right': {}, 'left': {}}
- for side in self.wings_skeleton3d_init.keys():
- for label in wings_params[obj_name][side].keys():
- wings_params_frame[side][label] = wings_params[obj_name][side][label][i]
-
- wings_skeleton3d_new = copy.deepcopy(self.wings_skeleton3d_init)
- wings_skeleton2d_new = {'right': {}, 'left': {}}
- for side in self.wings_skeleton3d_init.keys():
- wings_skeleton3d_new[side] = build_wing_skeleton3d(wings_skeleton3d_new[side], wings_params_frame[side])
- wings_skeleton3d_new[side] = \
- rotate_and_translate_wing_skeleton3d(wings_skeleton3d_new[side], wings_params_frame[side], side)
-
- if '2d' in res_method:
- wings_skeleton2d_new[side] = reproject_skeleton3d_to2d(wings_skeleton3d_new[side], self.dlt_coefs)
-
- for label in ['tip', 'hinge']:
- wing_label = side + '_wing_' + label
-
- if '2d' in res_method:
- for camn in range(1, self.nb_cam + 1):
- self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'][i] = wings_skeleton2d_new[side][label][camn][0]
- self.hybrid_skeleton2d[obj_name][wing_label][camn]['y'][i] = wings_skeleton2d_new[side][label][camn][1]
- self.hybrid_skeleton2d[obj_name][wing_label][camn]['likelihood'][i] = 1.0
-
- else:
- self.hybrid_skeleton3d[obj_name][wing_label]['x'][i] = wings_skeleton3d_new[side][label][0]
- self.hybrid_skeleton3d[obj_name][wing_label]['y'][i] = wings_skeleton3d_new[side][label][1]
- self.hybrid_skeleton3d[obj_name][wing_label]['z'][i] = wings_skeleton3d_new[side][label][2]
-
- print('>> first optim wing done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
-
- res_method = 'hybrid_' + res_method
-
- hybrid_param_ests, nb_visible_pts, rmse, nb_iterations = {}, {}, {}, {}
- obj_names = self.hybrid_skeleton2d['obj_names'] if '2d' in res_method else self.hybrid_skeleton3d['obj_names']
- for obj_name in obj_names:
- frames = self.hybrid_skeleton2d[obj_name]['frames'] if '2d' in res_method else self.hybrid_skeleton3d[obj_name]['frames']
- # frames = [frame for i, frame in enumerate(frames) if i % 100 == 0.0]
-
- # Low pass filter the position of wing tips and hinges + deviation_a and span
- for param_name in ['deviation_a', 'span']:
- hybrid_params_init[obj_name]['wbaverage_' + param_name] = \
- np.mean([np.transpose(filter_interp(wings_params_init[obj_name]['right'][param_name],
- self.framerate, self.cutoff_frequency)),
- np.transpose(filter_interp(wings_params_init[obj_name]['left'][param_name],
- self.framerate, self.cutoff_frequency))], axis=0)
-
- for side in self.wings_params_init.keys():
- for label in ['tip', 'hinge']:
- wing_label = side +'_wing_' + label
- if '2d' in res_method:
- for camn in range(1, self.nb_cam + 1):
- self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'] = \
- filter_interp(self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'],
- self.framerate, self.cutoff_frequency)
- self.hybrid_skeleton2d[obj_name][wing_label][camn]['y'] = \
- filter_interp(self.hybrid_skeleton2d[obj_name][wing_label][camn]['y'],
- self.framerate, self.cutoff_frequency)
- self.hybrid_skeleton2d[obj_name][wing_label][camn]['likelihood'] = \
- np.ones(np.size(self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'],
- self.framerate, self.cutoff_frequency))
-
- else:
- self.hybrid_skeleton3d[obj_name][wing_label]['x'] = \
- filter_interp(self.hybrid_skeleton3d[obj_name][wing_label]['x'],
- self.framerate, self.cutoff_frequency)
- self.hybrid_skeleton3d[obj_name][wing_label]['y'] = \
- filter_interp(self.hybrid_skeleton3d[obj_name][wing_label]['y'],
- self.framerate, self.cutoff_frequency)
- self.hybrid_skeleton3d[obj_name][wing_label]['z'] = \
- filter_interp(self.hybrid_skeleton3d[obj_name][wing_label]['z'],
- self.framerate, self.cutoff_frequency)
-
- shape_frame = np.array(frames).shape
- nb_iterations[obj_name] = np.zeros(shape_frame)
- nb_visible_pts[obj_name], rmse[obj_name] = np.ones(shape_frame) * np.nan, np.ones(shape_frame) * np.nan
- if obj_name in hybrid_params_init.keys():
- hybrid_param_ests[obj_name] = copy.deepcopy(hybrid_params_init[obj_name])
- else:
- hybrid_param_ests[obj_name] = {}
- for label in self.hybrid_params_init.keys():
- hybrid_param_ests[obj_name][label] = np.ones(shape_frame) * self.hybrid_params_init[label]
-
- if self.show_plot and not self.multiprocessing:
- fig, axs = plt.subplots(1, self.nb_cam)
- all_images = {}
- for camn in range (1, self.nb_cam +1):
- all_files = sorted(os.listdir(rec_paths[camn]), key=lambda x: x[20:20 + self.nb_leading_zero])
- all_images[camn] = [s for s in all_files if '.{0}'.format(self.image_format) in s]
-
- if self.multiprocessing: args = [(frame,) for frame in frames]
-
- #for i, frame in enumerate(tqdm.tqdm(frames)): # to print progress bar
- for i, frame in enumerate(frames):
- if obj_name in hybrid_params_init.keys():
- hybrid_params_init_frame = {}
- for label in self.hybrid_params_init.keys():
- hybrid_params_init_frame[label] = hybrid_params_init[obj_name][label][i]
- else:
- hybrid_params_init_frame = self.hybrid_params_init
-
- nb_points = len(self.hybrid_skeleton3d_init.keys())
- nb_visible_pts[obj_name][i] = nb_points
- if '3d' in res_method:
- hybrid_skeleton3d = {}
- for hybrid_label in self.hybrid_skeleton3d_init.keys():
- hybrid_skeleton3d[hybrid_label] = \
- np.array([self.hybrid_skeleton3d[obj_name][hybrid_label]['x'][i],
- self.hybrid_skeleton3d[obj_name][hybrid_label]['y'][i],
- self.hybrid_skeleton3d[obj_name][hybrid_label]['z'][i]])
-
- if np.isnan(hybrid_skeleton3d[hybrid_label]).any(): nb_visible_pts[obj_name][i] -= 1
-
- elif '2d' in res_method:
- hybrid_skeleton2d = {}
- for hybrid_label in self.hybrid_skeleton3d_init.keys():
- hybrid_skeleton2d[hybrid_label] = {1: np.array([])}
- for camn in range(1, self.nb_cam + 1):
-
- if self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['likelihood'][i] > self.threshold_likelihood:
- hybrid_skeleton2d[hybrid_label][camn] = \
- np.array([self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['x'][i],
- self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['y'][i],
- self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['likelihood'][i]])
- else:
- hybrid_skeleton2d[hybrid_label][camn] = \
- np.array([np.nan, np.nan, self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['likelihood'][i]])
-
- nb_visible_pts[obj_name][i] -= 1 / self.nb_cam
-
- return_nan = nb_visible_pts[obj_name][i] <= self.threshold_nb_pts_body - 2 # to compensate for the extra wing tips
-
- if not self.multiprocessing:
- if '3d' in res_method:
- hybrid_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
- optim_fit_body_params(animal_name, self.hybrid_skeleton3d_init, hybrid_skeleton3d, hybrid_params_init_frame,
- param_names, res_method, opt_method, self.hybrid_bounds_init, return_nan)
-
- elif '2d' in res_method:
- hybrid_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
- optim_fit_body_params(animal_name, self.hybrid_skeleton3d_init, hybrid_skeleton2d, hybrid_params_init_frame,
- param_names, res_method, opt_method, self.hybrid_bounds_init, return_nan, dlt_coefs=self.dlt_coefs)
-
- for label in hybrid_param_ests_frame.keys():
- hybrid_param_ests[obj_name][label][i] = hybrid_param_ests_frame[label]
-
- if self.show_plot and (frame % 20 == 0.0 or frame == 1):
- hybrid_skeleton3d_init = copy.deepcopy(self.hybrid_skeleton3d_init)
- hybrid_skeleton3d_init = build_hybrid_skeleton3d(hybrid_skeleton3d_init, hybrid_params_init_frame)
- hybrid_skeleton3d_init = rotate_body_skeleton3d(hybrid_skeleton3d_init, hybrid_params_init_frame)
- hybrid_skeleton3d_init = translate_body_skeleton3d(hybrid_skeleton3d_init, hybrid_params_init_frame)
- hybrid_skeleton2d_init = reproject_skeleton3d_to2d(hybrid_skeleton3d_init, self.dlt_coefs)
-
- hybrid_skeleton3d_new = copy.deepcopy(self.hybrid_skeleton3d_init)
- hybrid_skeleton3d_new = build_hybrid_skeleton3d(hybrid_skeleton3d_new, hybrid_param_ests_frame)
- hybrid_skeleton3d_new = rotate_body_skeleton3d(hybrid_skeleton3d_new, hybrid_param_ests_frame)
- hybrid_skeleton3d_new = translate_body_skeleton3d(hybrid_skeleton3d_new, hybrid_param_ests_frame)
- hybrid_skeleton2d_new = reproject_skeleton3d_to2d(hybrid_skeleton3d_new, self.dlt_coefs)
-
- # Compute reprojection of 3d reconstruction from dlc 2d points
- hybrid_skeleton3d_repro2d = {}
- for label in self.hybrid_skeleton3d_init.keys():
- hybrid_skeleton3d_repro2d[label] = {1: []}
- for j, camn in enumerate(range(1, self.nb_cam + 1)):
- uv = calib.find2d(1, self.dlt_coefs[j], np.array([[self.hybrid_skeleton3d[obj_name][label]['x'][i],
- self.hybrid_skeleton3d[obj_name][label]['y'][i], self.hybrid_skeleton3d[obj_name][label]['z'][i]]]))
- hybrid_skeleton3d_repro2d[label][camn] = np.array([uv[0][0], uv[0][1], 1.0])
-
- if '3d' in res_method:
- hybrid_skeleton2d = reproject_skeleton3d_to2d(hybrid_skeleton3d, self.dlt_coefs)
-
- # Plot 2d repojection of hybrid skeleton
- for j, camn in enumerate(range(1, self.nb_cam +1)):
- center_of_mass_xy = \
- (hybrid_skeleton2d_new['torso_abdomen_joint'][camn] * hybrid_param_ests_frame['ratio_com_torso'] \
- + hybrid_skeleton2d_new['head_torso_joint'][camn] * (1 - hybrid_param_ests_frame['ratio_com_torso']))
-
- if np.isnan(center_of_mass_xy).any(): continue
-
- axs[j].clear()
- axs[j].imshow(Image.open(os.path.join(rec_paths[camn], all_images[camn][i])))
- for label in hybrid_skeleton2d_new.keys():
- if hybrid_skeleton2d[label][camn][2] > self.threshold_likelihood:
- axs[j].scatter(hybrid_skeleton2d[label][camn][0],
- hybrid_skeleton2d[label][camn][1], marker='.', color='k')
- axs[j].scatter(hybrid_skeleton3d_repro2d[label][camn][0],
- hybrid_skeleton3d_repro2d[label][camn][1], marker='o', facecolors='none', edgecolors='k')
- axs[j].scatter(hybrid_skeleton2d_init[label][camn][0],
- hybrid_skeleton2d_init[label][camn][1], marker='o', facecolors='none', edgecolors='b')
- axs[j].scatter(hybrid_skeleton2d_new[label][camn][0],
- hybrid_skeleton2d_new[label][camn][1], marker='+', color='r')
-
- axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
- axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
-
- plt.pause(0.05)
- # plt.show()
-
- # param_to_print = ['pitch_a', 'roll_a', 'yaw_a', 'ratio_hybrid']
- # for label in param_to_print:
- # print('{0}: init: {1} and est: {2}'.format(label, hybrid_params_init_frame[label], hybrid_param_ests_frame[label]))
-
- else:
- if '3d' in res_method:
- args[i] = (self.hybrid_skeleton3d_init, hybrid_skeleton3d, hybrid_params_init_frame, param_names,
- res_method, opt_method, self.hybrid_bounds_init, return_nan,)
-
- elif '2d' in res_method:
- args[i] = (self.hybrid_skeleton3d_init, hybrid_skeleton2d, hybrid_params_init_frame, param_names,
- res_method, opt_method, self.hybrid_bounds_init, return_nan,)
-
- if self.multiprocessing:
- pool_hybrid = Pool(self.nb_process)
- if '3d' in res_method:
- results = pool_hybrid.starmap(optim_fit_body_params, args)
-
- elif '2d' in res_method:
- results = pool_hybrid.starmap(partial(optim_fit_body_params, dlt_coefs=self.dlt_coefs), args)
-
- pool_hybrid.close()
- for i, frame in enumerate(frames):
- hybrid_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = results[i]
- for label in hybrid_param_ests_frame.keys():
- hybrid_param_ests[obj_name][label][i] = hybrid_param_ests_frame[label]
-
- hybrid_param_ests[obj_name] = self.filter_interp_params(hybrid_param_ests[obj_name], param_names,
- interpolate_nans, filter_high_freq)
-
- return hybrid_param_ests, nb_visible_pts, rmse, nb_iterations
-
- def optim_fit_all_body_params(self, animal_name, body_params_init, param_names, res_method, opt_method, rec_paths,
- interpolate_nans=False, filter_high_freq=False):
-
- res_method = 'body_' + res_method
-
- body_param_ests, nb_visible_pts, rmse, nb_iterations = {}, {}, {}, {}
- obj_names = self.skeleton2d['obj_names'] if '2d' in res_method else self.skeleton3d['obj_names']
- for obj_name in obj_names:
- frames = self.skeleton2d[obj_name]['frames'] if '2d' in res_method else self.skeleton3d[obj_name]['frames']
- # frames = [frame for i, frame in enumerate(frames) if i % 100 == 0.0]
-
- shape_frame = np.array(frames).shape
- nb_iterations[obj_name] = np.zeros(shape_frame)
- rmse[obj_name], nb_visible_pts[obj_name] = np.ones(shape_frame) * np.nan, np.ones(shape_frame) * np.nan
- if obj_name in body_params_init.keys():
- body_param_ests[obj_name] = copy.deepcopy(body_params_init[obj_name])
- else:
- body_param_ests[obj_name] = {}
- for label in self.body_params_init.keys():
- body_param_ests[obj_name][label] = np.ones(shape_frame) * self.body_params_init[label]
-
- if self.show_plot and not self.multiprocessing:
- fig, axs = plt.subplots(1, self.nb_cam)
- all_images = {}
- for camn in range (1, self.nb_cam +1):
- all_files = sorted(os.listdir(rec_paths[camn]), key=lambda x: x[20:20 + self.nb_leading_zero])
- all_images[camn] = [s for s in all_files if '.{0}'.format(self.image_format) in s]
-
- if self.multiprocessing: args = [(frame,) for frame in frames]
-
- #for i, frame in enumerate(tqdm.tqdm(frames)): # to print progress bar
- for i, frame in enumerate(frames):
- if obj_name in body_params_init.keys():
- body_params_init_frame = {}
- for label in self.body_params_init.keys():
- body_params_init_frame[label] = body_params_init[obj_name][label][i]
- else:
- body_params_init_frame = self.body_params_init
-
- nb_points = len(self.body_skeleton3d_init.keys())
- nb_visible_pts[obj_name][i] = nb_points
- if '3d' in res_method:
- body_skeleton3d = {}
- for body_label in self.body_skeleton3d_init.keys():
- body_skeleton3d[body_label] = \
- np.array([self.skeleton3d[obj_name][body_label]['x'][i],
- self.skeleton3d[obj_name][body_label]['y'][i],
- self.skeleton3d[obj_name][body_label]['z'][i]])
-
- if np.isnan(body_skeleton3d[body_label]).any(): nb_visible_pts[obj_name][i] -= 1
-
- elif '2d' in res_method:
- body_skeleton2d = {}
- for body_label in self.body_skeleton3d_init.keys():
- body_skeleton2d[body_label] = {1: np.array([])}
- for camn in range(1, self.nb_cam + 1):
-
- if self.skeleton2d[obj_name][body_label][camn]['likelihood'][i] > self.threshold_likelihood:
- body_skeleton2d[body_label][camn] = \
- np.array([self.skeleton2d[obj_name][body_label][camn]['x'][i],
- self.skeleton2d[obj_name][body_label][camn]['y'][i],
- self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
- else:
- body_skeleton2d[body_label][camn] = \
- np.array([np.nan, np.nan, self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
-
- nb_visible_pts[obj_name][i] -= 1 / self.nb_cam
-
- return_nan = nb_visible_pts[obj_name][i] <= self.threshold_nb_pts_body
-
- if not self.multiprocessing:
- if '3d' in res_method:
- body_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
- optim_fit_body_params(animal_name, self.body_skeleton3d_init, body_skeleton3d, body_params_init_frame,
- param_names, res_method, opt_method, self.body_bounds_init, return_nan)
-
- elif '2d' in res_method:
- body_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
- optim_fit_body_params(animal_name, self.body_skeleton3d_init, body_skeleton2d, body_params_init_frame,
- param_names, res_method, opt_method, self.body_bounds_init, return_nan, dlt_coefs=self.dlt_coefs)
-
- for label in body_param_ests_frame.keys():
- body_param_ests[obj_name][label][i] = body_param_ests_frame[label]
-
- if self.show_plot and (frame % 20 == 0.0 or frame == 1):
- body_skeleton3d_init = copy.deepcopy(self.body_skeleton3d_init)
- body_skeleton3d_init = build_body_skeleton3d(body_skeleton3d_init, body_params_init_frame)
- body_skeleton3d_init = rotate_body_skeleton3d(body_skeleton3d_init, body_params_init_frame)
- body_skeleton3d_init = translate_body_skeleton3d(body_skeleton3d_init, body_params_init_frame)
- body_skeleton2d_init = reproject_skeleton3d_to2d(body_skeleton3d_init, self.dlt_coefs)
-
- body_skeleton3d_new = copy.deepcopy(self.body_skeleton3d_init)
- body_skeleton3d_new = build_body_skeleton3d(body_skeleton3d_new, body_param_ests_frame)
- body_skeleton3d_new = rotate_body_skeleton3d(body_skeleton3d_new, body_param_ests_frame)
- body_skeleton3d_new = translate_body_skeleton3d(body_skeleton3d_new, body_param_ests_frame)
- body_skeleton2d_new = reproject_skeleton3d_to2d(body_skeleton3d_new, self.dlt_coefs)
-
- # Compute reprojection of 3d reconstruction from dlc 2d points
- body_skeleton3d_repro2d = {}
- for label in self.body_skeleton3d_init.keys():
- body_skeleton3d_repro2d[label] = {1: []}
- for j, camn in enumerate(range(1, self.nb_cam + 1)):
- uv = calib.find2d(1, self.dlt_coefs[j], np.array([[self.skeleton3d[obj_name][label]['x'][i],
- self.skeleton3d[obj_name][label]['y'][i], self.skeleton3d[obj_name][label]['z'][i]]]))
- body_skeleton3d_repro2d[label][camn] = np.array([uv[0][0], uv[0][1], 1.0])
-
- if '3d' in res_method:
- body_skeleton2d = reproject_skeleton3d_to2d(body_skeleton3d, self.dlt_coefs)
-
- # Plot 2d repojection of body skeleton
- for j, camn in enumerate(range(1, self.nb_cam +1)):
- center_of_mass_xy = \
- (body_skeleton2d_new['torso_abdomen_joint'][camn] * body_param_ests_frame['ratio_com_torso'] \
- + body_skeleton2d_new['head_torso_joint'][camn] * (1 - body_param_ests_frame['ratio_com_torso']))
-
- if np.isnan(center_of_mass_xy).any(): continue
-
- axs[j].clear()
- axs[j].imshow(Image.open(os.path.join(rec_paths[camn], all_images[camn][i])))
- for label in body_skeleton2d_new.keys():
- if body_skeleton2d[label][camn][2] > self.threshold_likelihood:
- axs[j].scatter(body_skeleton2d[label][camn][0],
- body_skeleton2d[label][camn][1], marker='.', color='k')
- axs[j].scatter(body_skeleton3d_repro2d[label][camn][0],
- body_skeleton3d_repro2d[label][camn][1], marker='o', facecolors='none', edgecolors='k')
- axs[j].scatter(body_skeleton2d_init[label][camn][0],
- body_skeleton2d_init[label][camn][1], marker='o', facecolors='none', edgecolors='b')
- axs[j].scatter(body_skeleton2d_new[label][camn][0],
- body_skeleton2d_new[label][camn][1], marker='+', color='r')
-
- axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
- axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
-
- plt.pause(0.05)
- # plt.show()
-
- # param_to_print = ['pitch_a', 'roll_a', 'yaw_a', 'ratio_body']
- # for label in param_to_print:
- # print('{0}: init: {1} and est: {2}'.format(label, body_params_init_frame[label], body_param_ests_frame[label]))
-
- else:
- if '3d' in res_method:
- args[i] = (self.body_skeleton3d_init, body_skeleton3d, body_params_init_frame, param_names,
- res_method, opt_method, self.body_bounds_init, return_nan,)
-
- elif '2d' in res_method:
- args[i] = (self.body_skeleton3d_init, body_skeleton2d, body_params_init_frame, param_names,
- res_method, opt_method, self.body_bounds_init, return_nan,)
-
- if self.multiprocessing:
- pool_body = Pool(self.nb_process)
- if '3d' in res_method:
- results = pool_body.starmap(optim_fit_body_params, args)
-
- elif '2d' in res_method:
- results = pool_body.starmap(partial(optim_fit_body_params, dlt_coefs=self.dlt_coefs), args)
-
- pool_body.close()
- for i, frame in enumerate(frames):
- body_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = results[i]
- for label in body_param_ests_frame.keys():
- body_param_ests[obj_name][label][i] = body_param_ests_frame[label]
-
- body_param_ests[obj_name] = self.filter_interp_params(body_param_ests[obj_name], param_names,
- interpolate_nans, filter_high_freq)
-
- return body_param_ests, nb_visible_pts, rmse, nb_iterations
-
- def optim_fit_all_wings_params(self, animal_name, body_params_init, wings_params_init, param_names, res_method, opt_method,
- rec_paths, interpolate_nans=False, filter_high_freq=False):
-
- res_method = 'wing_' + res_method
-
- obj_names = self.skeleton2d['obj_names'] if '2d' in res_method else self.skeleton3d['obj_names']
- for obj_name in obj_names:
- if obj_name in wings_params_init.keys():
- for side in self.wings_params_init.keys():
- for label in body_params_init[obj_name].keys():
- wings_params_init[obj_name][side][label] = body_params_init[obj_name][label]
-
- wings_params_ests, nb_visible_pts, rmse, nb_iterations = copy.deepcopy(wings_params_init), {}, {}, {}
- for obj_name in obj_names:
- frames = self.skeleton2d[obj_name]['frames'] if '2d' in res_method else self.skeleton3d[obj_name]['frames']
- # frames = [frame for i, frame in enumerate(frames) if i % 100 == 0.0]
-
- if self.show_plot and not self.multiprocessing:
- fig, axs = plt.subplots(1, self.nb_cam)
- all_images = {}
- for camn in range(1, self.nb_cam + 1):
- all_files = sorted(os.listdir(rec_paths[camn]), key=lambda x: x[20:20 + self.nb_leading_zero])
- all_images[camn] = [s for s in all_files if '.{0}'.format(self.image_format) in s]
-
- if self.multiprocessing: args = [(frame,) for frame in frames]
-
- shape_frame = np.array(frames).shape
- rmse[obj_name] = {'right': np.ones(shape_frame) * np.nan, 'left': np.ones(shape_frame) * np.nan}
- nb_iterations[obj_name] = {'right': np.zeros(shape_frame), 'left': np.zeros(shape_frame)}
- nb_visible_pts[obj_name] = {'right': np.ones(shape_frame) * np.nan, 'left': np.ones(shape_frame) * np.nan}
- #for i, frame in enumerate(tqdm.tqdm(frames)): # to print progress bar
- for i, frame in enumerate(frames):
- if obj_name in wings_params_init.keys():
- wings_params_init_frame = {}
- for side in self.wings_params_init.keys():
- wings_params_init_frame[side] = {}
- for label in self.wings_params_init[side].keys():
- wings_params_init_frame[side][label] = wings_params_init[obj_name][side][label][i]
- else:
- wings_params_init_frame = self.wings_params_init
-
- nb_points = len(self.body_skeleton3d_init.keys())
- for side in self.wings_skeleton3d_init.keys():
- nb_visible_pts[obj_name][side][i] = nb_points
-
- if '3d' in res_method:
- wings_skeleton3d = {'right': {}, 'left': {}}
- for side in self.wings_skeleton3d_init.keys():
- for wing_label in self.wings_skeleton3d_init[side].keys():
- label = side + '_wing_' + wing_label
- wings_skeleton3d[side][wing_label] = \
- np.array([self.skeleton3d[obj_name][label]['x'][i],
- self.skeleton3d[obj_name][label]['y'][i],
- self.skeleton3d[obj_name][label]['z'][i]])
-
- if np.isnan(wings_skeleton3d[side][wing_label]).any():
- nb_visible_pts[obj_name][side][i] -= 1
-
- elif '2d' in res_method:
- wings_skeleton2d = {'right': {}, 'left': {}}
- for side in self.wings_skeleton3d_init.keys():
- for wing_label in self.wings_skeleton3d_init[side].keys():
-
- wings_skeleton2d[side][wing_label] = {1: []}
- for camn in range(1, self.nb_cam + 1):
- label = side + '_wing_' + wing_label
-
- if self.skeleton2d[obj_name][label][camn]['likelihood'][i] > self.threshold_likelihood:
- wings_skeleton2d[side][wing_label][camn] = \
- np.array([self.skeleton2d[obj_name][label][camn]['x'][i],
- self.skeleton2d[obj_name][label][camn]['y'][i],
- self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
- else:
- wings_skeleton2d[side][wing_label][camn] = \
- np.array([np.nan, np.nan, self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
-
- nb_visible_pts[obj_name][side][i] -= 1 / self.nb_cam
-
- body_params_frame = {}
- for label in body_params_init[obj_name].keys():
- body_params_frame[label] = body_params_init[obj_name][label][i]
-
- body_skeleton3d = copy.deepcopy(self.body_skeleton3d_init)
- body_skeleton3d = build_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton3d = rotate_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton3d = translate_body_skeleton3d(body_skeleton3d, body_params_frame)
-
- return_nan = [False, False]
- for j, side in enumerate(self.wings_skeleton3d_init.keys()):
- [wings_params_init_frame[side]['x_hinge'], wings_params_init_frame[side]['y_hinge'],
- wings_params_init_frame[side]['z_hinge']] = body_skeleton3d['{0}_wing_hinge'.format(side)]
-
- return_nan[j] = nb_visible_pts[obj_name][side][i] <= self.threshold_nb_pts_wing
- if return_nan[j]:
- for label in param_names:
- wings_params_ests[obj_name][side][label][i] = np.nan
- continue
-
- for label in ['yaw_a', 'pitch_a', 'roll_a', 'x_com', 'y_com', 'z_com']:
- wings_params_init_frame[side][label] = body_params_frame[label]
-
- if not self.multiprocessing:
- if 'geo' in res_method:
- if '3d' in res_method:
- wings_params_est_frame, rmse_frame, nb_iterations_frame = \
- optim_fit_limbs_params(animal_name, self.wings_geometry3d_init, wings_skeleton3d, wings_params_init_frame,
- param_names, res_method, opt_method, self.wing_bounds_init, return_nan)
-
- elif '2d' in res_method:
- wings_params_est_frame, rmse_frame, nb_iterations_frame = \
- optim_fit_limbs_params(animal_name, self.wings_geometry3d_init, wings_skeleton2d, wings_params_init_frame,
- param_names, res_method, opt_method, self.wing_bounds_init, return_nan,
- dlt_coefs=self.dlt_coefs)
- else:
- if '3d' in res_method:
- wings_params_est_frame, rmse_frame, nb_iterations_frame = \
- optim_fit_limbs_params(animal_name, self.wings_skeleton3d_init, wings_skeleton3d, wings_params_init_frame,
- param_names, res_method, opt_method, self.wing_bounds_init, return_nan)
-
- elif '2d' in res_method:
- wings_params_est_frame, rmse_frame, nb_iterations_frame = \
- optim_fit_limbs_params(animal_name, self.wings_skeleton3d_init, wings_skeleton2d, wings_params_init_frame,
- param_names, res_method, opt_method, self.wing_bounds_init, return_nan,
- dlt_coefs=self.dlt_coefs)
-
- for side in self.wings_skeleton3d_init.keys():
- rmse[obj_name][side][i] = rmse_frame[side]
- nb_iterations[obj_name][side][i] = nb_iterations_frame[side]
- for label in wings_params_est_frame[side].keys():
- wings_params_ests[obj_name][side][label][i] = wings_params_est_frame[side][label]
-
- if self.show_plot and frame % 20 == 0.0:
- body_skeleton2d_new = reproject_skeleton3d_to2d(body_skeleton3d, self.dlt_coefs)
-
- wings_skeleton3d_new = copy.deepcopy(self.wings_skeleton3d_init)
- wings_skeleton3d_init = copy.deepcopy(self.wings_skeleton3d_init)
- wings_skeleton2d_new, wings_skeleton2d_init = {'right': {}, 'left': {}}, {'right': {}, 'left': {}}
- wings_skeleton3d_repro2d = {'right': {}, 'left': {}}
- if '3d' in res_method: wings_skeleton2d = copy.deepcopy(wings_skeleton2d_new)
- for side in self.wings_skeleton3d_init.keys():
- wings_skeleton3d_init[side] = build_wing_skeleton3d(wings_skeleton3d_init[side], wings_params_init_frame[side])
- wings_skeleton3d_init[side] = rotate_and_translate_wing_skeleton3d(wings_skeleton3d_init[side], wings_params_init_frame[side], side)
- wings_skeleton2d_init[side] = reproject_skeleton3d_to2d(wings_skeleton3d_init[side], self.dlt_coefs)
-
- wings_skeleton3d_new[side] = build_wing_skeleton3d(wings_skeleton3d_new[side], wings_params_est_frame[side])
- wings_skeleton3d_new[side] = rotate_and_translate_wing_skeleton3d(wings_skeleton3d_new[side], wings_params_est_frame[side], side)
- wings_skeleton2d_new[side] = reproject_skeleton3d_to2d(wings_skeleton3d_new[side], self.dlt_coefs)
-
- # Compute reprojection of 3d reconstruction from dlc 2d points
- for label in self.wings_skeleton3d_init[side].keys():
- wings_skeleton3d_repro2d[side][label] = {1: []}
- for j, camn in enumerate(range(1, self.nb_cam + 1)):
- label2 = side + '_wing_' + label
- uv = calib.find2d(1, self.dlt_coefs[j], np.array([[self.skeleton3d[obj_name][label2]['x'][i],
- self.skeleton3d[obj_name][label2]['y'][i], self.skeleton3d[obj_name][label2]['z'][i]]]))
- wings_skeleton3d_repro2d[side][label][camn] = np.array([uv[0][0], uv[0][1], 1.0])
-
- if '3d' in res_method:
- wings_skeleton2d[side] = reproject_skeleton3d_to2d(wings_skeleton3d[side], self.dlt_coefs)
-
- for j, camn in enumerate(range(1, self.nb_cam + 1)):
- center_of_mass_xy = \
- (body_skeleton2d_new['torso_abdomen_joint'][camn] * body_params_frame['ratio_com_torso'] \
- + body_skeleton2d_new['head_torso_joint'][camn] * (1 - body_params_frame['ratio_com_torso']))
-
- if np.isnan(center_of_mass_xy).any(): continue
-
- axs[j].clear()
- axs[j].imshow(Image.open(os.path.join(rec_paths[camn], all_images[camn][i])))
- for side in self.wings_skeleton3d_init.keys():
- for label in wings_skeleton2d_new[side].keys():
- if wings_skeleton2d[side][label][camn][2] > self.threshold_likelihood:
- axs[j].scatter(wings_skeleton2d[side][label][camn][0],
- wings_skeleton2d[side][label][camn][1], marker='.', color='k')
- axs[j].scatter(wings_skeleton3d_repro2d[side][label][camn][0],
- wings_skeleton3d_repro2d[side][label][camn][1], marker='o', facecolors='none', edgecolors='k')
- axs[j].scatter(wings_skeleton2d_init[side][label][camn][0],
- wings_skeleton2d_init[side][label][camn][1], marker='o', facecolors='none', edgecolors='b')
- axs[j].scatter(wings_skeleton2d_new[side][label][camn][0],
- wings_skeleton2d_new[side][label][camn][1], marker='+', color='r')
-
- axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
- axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
-
- plt.pause(0.05)
- #plt.show()
-
- # param_to_print = ['stroke_a', 'deviation_a', 'rotation_a', 'ratio_wing']
- # for label in param_to_print:
- # print('{0}: init: {1} and est: {2}'.format(label, wings_params_init_frame[side][label],
- # wings_params_est_frame[side][label]))
-
- else:
- if 'geo' in res_method:
- if '3d' in res_method:
- args[i] = (self.wings_geometry3d_init, wings_skeleton3d, wings_params_init_frame, param_names,
- res_method, opt_method, self.wing_bounds_init, return_nan)
-
- elif '2d' in res_method:
- args[i] = (self.wings_geometry3d_init, wings_skeleton2d, wings_params_init_frame, param_names,
- res_method, opt_method, self.wing_bounds_init, return_nan)
- else:
- if '3d' in res_method:
- args[i] = (self.wings_skeleton3d_init, wings_skeleton3d, wings_params_init_frame, param_names,
- res_method, opt_method, self.wing_bounds_init, return_nan)
-
- elif '2d' in res_method:
- args[i] = (self.wings_skeleton3d_init, wings_skeleton2d, wings_params_init_frame, param_names,
- res_method, opt_method, self.wing_bounds_init, return_nan)
-
- if self.multiprocessing:
- pool_wings = Pool(self.nb_process)
- if '3d' in res_method:
- results = pool_wings.starmap(optim_fit_limbs_params, args)
-
- elif '2d' in res_method:
- results = pool_wings.starmap(partial(optim_fit_limbs_params, dlt_coefs=self.dlt_coefs), args)
-
- pool_wings.close()
- for i, frame in enumerate(frames):
- wings_params_est_frame, rmse_frame, nb_iterations_frame = results[i]
- for side in self.wings_skeleton3d_init.keys():
- rmse[obj_name][side][i] = rmse_frame[side]
- nb_iterations[obj_name][side][i] = nb_iterations_frame[side]
- for label in wings_params_est_frame[side].keys():
- wings_params_ests[obj_name][side][label][i] = wings_params_est_frame[side][label]
-
- for side in self.wings_skeleton3d_init.keys():
- wings_params_ests[obj_name][side] = \
- self.filter_interp_params(wings_params_ests[obj_name][side], ['x_hinge', 'y_hinge', 'z_hinge'],
- False, filter_high_freq)
- wings_params_ests[obj_name][side] = \
- self.filter_interp_params(wings_params_ests[obj_name][side], param_names, interpolate_nans, False)
-
- return wings_params_ests, nb_visible_pts, rmse, nb_iterations
-
- def correct_body_wings_params(self, body_params, wings_params, angle_names_to_correct, rec_paths):
- # Will correct asymetrical wing angles (mean value should be the same on both side)
- body_params_corrected = copy.deepcopy(body_params)
- wings_params_corrected = copy.deepcopy(wings_params)
-
- obj_names = self.skeleton2d['obj_names']
- for obj_name in obj_names:
-
- wings_params_mean = {} # Find angles error (btw stroke_a and deviation_a of both sides)
- for side in self.wings_skeleton3d_init.keys():
- wings_params_mean[side] = \
- self.filter_interp_params(wings_params[obj_name][side], angle_names_to_correct, False, True)
-
- if 'stroke_a' in angle_names_to_correct:
- diff_stroke_a = wings_params_mean['right']['stroke_a'] - wings_params_mean['left']['stroke_a']
- # body_params_corrected[obj_name]['yaw_a'] -= diff_stroke_a / 2
- # wings_params_corrected[obj_name]['right']['stroke_a'] -= diff_stroke_a / 2
- # wings_params_corrected[obj_name]['left']['stroke_a'] += diff_stroke_a / 2
- # wings_params_corrected[obj_name]['right']['yaw_a'] -= diff_stroke_a / 2
- # wings_params_corrected[obj_name]['left']['yaw_a'] -= diff_stroke_a / 2
-
- if 'deviation_a' in angle_names_to_correct:
- diff_deviation_a = wings_params_mean['right']['deviation_a'] - wings_params_mean['left']['deviation_a']
- # body_params_corrected[obj_name]['roll_a'] -= diff_deviation_a / 2
- # wings_params_corrected[obj_name]['right']['deviation_a'] -= diff_deviation_a / 2
- # wings_params_corrected[obj_name]['left']['deviation_a'] += diff_deviation_a / 2
- # wings_params_corrected[obj_name]['right']['roll_a'] -= diff_deviation_a / 2
- # wings_params_corrected[obj_name]['left']['roll_a'] -= diff_deviation_a / 2
-
- if 'rotation_a' in angle_names_to_correct:
- raise ValueError('There is not reason for rotation_a to be corrected!'
- 'Any asymmetry btw sides cannot results from estimation error of the pitch angle')
-
- if self.show_plot:
- # Plot wing tips pos
- frames = self.skeleton2d[obj_name]['frames']
- wing_tips, wing_hinges = {'right': {}, 'left': {}}, {'right': {}, 'left': {}}
- for side in self.wings_skeleton3d_init.keys():
- for coord in ['x', 'y', 'z']:
- wing_tips[side][coord], wing_hinges[side][coord] = np.ones(np.size(frames)) * np.nan, np.ones(
- np.size(frames)) * np.nan
-
- tip_hinge_a = np.ones(np.size(frames)) * np.nan
- tip_torso_a = np.ones(np.size(frames)) * np.nan
- for i, frame in enumerate(frames):
- body_params_frame, wings_params_frame = {}, {'right': {}, 'left': {}}
- for label in body_params[obj_name].keys():
- body_params_frame[label] = body_params[obj_name][label][i]
-
- body_skeleton3d_est = copy.deepcopy(self.body_skeleton3d_init)
- body_skeleton3d_est = build_body_skeleton3d(body_skeleton3d_est, body_params_frame)
- body_skeleton3d_est = rotate_body_skeleton3d(body_skeleton3d_est, body_params_frame)
- body_skeleton3d_est = translate_body_skeleton3d(body_skeleton3d_est, body_params_frame)
-
- wings_skeleton3d_mean = copy.deepcopy(self.wings_skeleton3d_init)
- for side in self.wings_skeleton3d_init.keys():
- for label in wings_params[obj_name][side].keys():
- wings_params_frame[side][label] = wings_params_mean[side][label][i]
-
- wings_skeleton3d_mean[side] = build_wing_skeleton3d(wings_skeleton3d_mean[side],
- wings_params_frame[side])
- wings_skeleton3d_mean[side] = rotate_and_translate_wing_skeleton3d(wings_skeleton3d_mean[side],
- wings_params_frame[side],
- side)
-
- wing_tips[side]['x'][i], wing_tips[side]['y'][i], wing_tips[side]['z'][i] = \
- wings_skeleton3d_mean[side]['tip'][0], wings_skeleton3d_mean[side]['tip'][1], \
- wings_skeleton3d_mean[side]['tip'][2]
- wing_hinges[side]['x'][i], wing_hinges[side]['y'][i], wing_hinges[side]['z'][i] = \
- wings_skeleton3d_mean[side]['hinge'][0], wings_skeleton3d_mean[side]['hinge'][1], \
- wings_skeleton3d_mean[side]['hinge'][2]
-
- tip_vect = np.array(
- [wing_tips['right']['x'][i], wing_tips['right']['y'][i], wing_tips['right']['z'][i]]) \
- - np.array(
- [wing_tips['left']['x'][i], wing_tips['left']['y'][i], wing_tips['left']['z'][i]])
- hinge_vect = np.array(
- [wing_hinges['right']['x'][i], wing_hinges['right']['y'][i], wing_hinges['right']['z'][i]]) \
- - np.array(
- [wing_hinges['left']['x'][i], wing_hinges['left']['y'][i], wing_hinges['left']['z'][i]])
-
- torso_vect = body_skeleton3d_est['head_torso_joint'] - body_skeleton3d_est[
- 'torso_abdomen_joint']
-
- tip_hinge_a[i] = angle_from_vectors(tip_vect, hinge_vect)
- tip_torso_a[i] = angle_from_vectors(tip_vect, torso_vect)
-
- fig_test, axs_test = plt.subplots(1, 3)
- fig_test.suptitle('Average position of wing tip over time')
- for side in self.wings_skeleton3d_init.keys():
- axs_test[0].plot(wing_tips[side]['x'] - body_params[obj_name]['x_com'], label=side + '_tip')
- axs_test[1].plot(wing_tips[side]['y'] - body_params[obj_name]['y_com'], label=side + '_tip')
- axs_test[2].plot(wing_tips[side]['z'] - body_params[obj_name]['z_com'], label=side + '_tip')
-
- axs_test[0].plot(wing_hinges[side]['x'] - body_params[obj_name]['x_com'], label=side + '_hinge')
- axs_test[1].plot(wing_hinges[side]['y'] - body_params[obj_name]['y_com'], label=side + '_hinge')
- axs_test[2].plot(wing_hinges[side]['z'] - body_params[obj_name]['z_com'], label=side + '_hinge')
-
- axs_test[0].set_ylabel('x (m)'), axs_test[1].set_ylabel('y (m)'), axs_test[2].set_ylabel('z (m)')
-
- axs_test[2].set_xlabel('frames')
- fig_test.legend()
-
- fig_test2, axs_test2 = plt.subplots(1, 2)
- fig_test2.suptitle('Average angle btw wing tip vector and torso vector over time')
- axs_test2[0].plot(tip_hinge_a, label='tip_hinge_a')
- axs_test2[0].plot(diff_stroke_a, label='diff_stroke_a')
- axs_test2[0].plot(diff_deviation_a, label='diff_deviation_a')
- # axs_test2[0].plot(diff_rotation_a, label='diff_rotation_a')
- axs_test2[1].plot(tip_torso_a, label='tip_torso_a')
- axs_test2[0].set_ylabel('angle (deg)')
-
- fig_test2.legend()
-
- self.plot_skeleton2d(body_params_corrected, wings_params_corrected, rec_paths)
-
- return body_params_corrected, wings_params_corrected
-
- def plot_skeleton2d(self, body_params, wings_params, rec_paths, modulo=1,
- show_body=True, show_wings=True, save_images=False, save_paths={}):
-
- # TODO remove the folowing data in code
- camns_to_rotate = [2]
- degrees = 270
-
- obj_names = self.skeleton2d['obj_names']
- for obj_name in obj_names:
- if not show_body and not show_wings: break
-
- if self.show_plot:
- fig, axs = plt.subplots(1, self.nb_cam, figsize=(15, 20))
- plt.axis('off')
-
- all_images = {}
- for camn in range(1, self.nb_cam + 1):
- all_files = sorted(os.listdir(rec_paths[camn]), key=lambda x: x[20:20 + self.nb_leading_zero])
- all_images[camn] = [s for s in all_files if '.{0}'.format(self.image_format) in s]
-
- frames = self.skeleton2d[obj_name]['frames']
- for i, frame in enumerate(frames):
- if not frame % modulo == 0.0 and not frame == 1: continue
-
- body_skeleton2d_dlc = {}
- for body_label in self.skeleton2d[obj_name]['label_names']:
- body_skeleton2d_dlc[body_label] = {1: np.array([])}
- for camn in range(1, self.nb_cam + 1):
- if self.skeleton2d[obj_name][body_label][camn]['likelihood'][i] > self.threshold_likelihood:
- body_skeleton2d_dlc[body_label][camn] = \
- np.array([self.skeleton2d[obj_name][body_label][camn]['x'][i],
- self.skeleton2d[obj_name][body_label][camn]['y'][i],
- self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
- else:
- body_skeleton2d_dlc[body_label][camn] = \
- np.array([np.nan, np.nan, self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
-
- if show_body:
- body_params_frame = {}
- for label in body_params[obj_name].keys():
- body_params_frame[label] = body_params[obj_name][label][i]
-
- body_skeleton3d = copy.deepcopy(self.body_skeleton3d_init)
- body_skeleton3d = build_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton3d = rotate_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton3d = translate_body_skeleton3d(body_skeleton3d, body_params_frame)
- body_skeleton2d = reproject_skeleton3d_to2d(body_skeleton3d, self.dlt_coefs)
-
- if show_wings:
- wings_skeleton3d = copy.deepcopy(self.wings_skeleton3d_init)
-
- wings_params_frame = {'right': {}, 'left': {}}
- wings_skeleton2d, wings_skeleton2d_dlc = {'right': {}, 'left': {}}, {'right': {}, 'left': {}}
- for side in self.wings_skeleton3d_init.keys():
- for label in wings_params[obj_name][side].keys():
- wings_params_frame[side][label] = wings_params[obj_name][side][label][i]
-
- wings_skeleton3d[side] = build_wing_skeleton3d(wings_skeleton3d[side], wings_params_frame[side])
- wings_skeleton3d[side] = rotate_and_translate_wing_skeleton3d(wings_skeleton3d[side], wings_params_frame[side], side)
- wings_skeleton2d[side] = reproject_skeleton3d_to2d(wings_skeleton3d[side], self.dlt_coefs)
-
- for wing_label in self.wings_skeleton3d_init[side].keys():
- wings_skeleton2d_dlc[side][wing_label] = {1: []}
- for camn in range(1, self.nb_cam + 1):
- label = side + '_wing_' + wing_label
- if self.skeleton2d[obj_name][label][camn]['likelihood'][i] > self.threshold_likelihood:
- wings_skeleton2d_dlc[side][wing_label][camn] = \
- np.array([self.skeleton2d[obj_name][label][camn]['x'][i],
- self.skeleton2d[obj_name][label][camn]['y'][i],
- self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
- else:
- wings_skeleton2d_dlc[side][wing_label][camn] = \
- np.array([np.nan, np.nan, self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
-
- prev_img = None
- for j, camn in enumerate(range(1, self.nb_cam + 1)):
- center_of_mass_xy = \
- (body_skeleton2d['torso_abdomen_joint'][camn] * body_params_frame['ratio_com_torso'] \
- + body_skeleton2d['head_torso_joint'][camn] * (1 - body_params_frame['ratio_com_torso']))
-
- if np.isnan(center_of_mass_xy).any(): continue
-
- if self.show_plot or save_images:
- img = read_image(os.path.join(rec_paths[camn], all_images[camn][i]))
-
- if self.show_plot:
- axs[j].clear()
- axs[j].imshow(img)
-
- if show_body:
- for label in body_skeleton2d.keys():
- if body_skeleton2d_dlc[label][camn][2] > self.threshold_likelihood:
- axs[j].scatter(body_skeleton2d_dlc[label][camn][0],
- body_skeleton2d_dlc[label][camn][1], marker='.', color='k')
- axs[j].scatter(body_skeleton2d[label][camn][0],
- body_skeleton2d[label][camn][1], marker='+', color='r')
-
- if show_wings:
- for side in self.wings_skeleton3d_init.keys():
- for label in wings_skeleton2d[side].keys():
- if wings_skeleton2d_dlc[side][label][camn][2] > self.threshold_likelihood:
- axs[j].scatter(wings_skeleton2d_dlc[side][label][camn][0],
- wings_skeleton2d_dlc[side][label][camn][1], marker='.', color='k')
- axs[j].scatter(wings_skeleton2d[side][label][camn][0],
- wings_skeleton2d[side][label][camn][1], marker='+', color='r')
-
- axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
- axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
-
- if camn == self.nb_cam:
- plt.pause(0.05)
- # fig.savefig(os.path.join(save_paths[self.main_camn], all_images[camn][i]),
- # bbox_inches='tight', transparent=True, pad_inches=0)
-
- if save_images:
- img = img.convert('RGB')
- draw = ImageDraw.Draw(img)
- if show_body:
- body_line = []
- for label in body_skeleton2d.keys():
- if body_skeleton2d_dlc[label][camn][2] > self.threshold_likelihood:
- draw.point((body_skeleton2d_dlc[label][camn][0], body_skeleton2d_dlc[label][camn][1]), fill='blue')
- draw.point((body_skeleton2d[label][camn][0], body_skeleton2d[label][camn][1]), fill='red')
-
- if not 'hinge' in label:
- body_line.append((body_skeleton2d[label][camn][0], body_skeleton2d[label][camn][1]))
- draw.line(body_line, fill='red', width=2)
-
- if show_wings:
- for side in self.wings_skeleton3d_init.keys():
- wing_line = []
- for label in wings_skeleton2d[side].keys():
- if wings_skeleton2d_dlc[side][label][camn][2] > self.threshold_likelihood:
- draw.point((wings_skeleton2d_dlc[side][label][camn][0], wings_skeleton2d_dlc[side][label][camn][1]), fill='blue')
-
- draw.point((wings_skeleton2d[side][label][camn][0], wings_skeleton2d[side][label][camn][1]), fill='red')
- wing_line.append((wings_skeleton2d[side][label][camn][0], wings_skeleton2d[side][label][camn][1]))
-
- frst_label = list(wings_skeleton2d[side].keys())[0]
- wing_line.append((wings_skeleton2d[side][frst_label][camn][0], wings_skeleton2d[side][frst_label][camn][1]))
- draw.line(wing_line, fill='red', width=2)
-
- left, right = int(center_of_mass_xy[0] - 100), int(center_of_mass_xy[0] + 100)
- top, bottom = int(center_of_mass_xy[1] - 100), int(center_of_mass_xy[1] + 100)
- img = img.crop((left, top, right, bottom))
-
- if camn in camns_to_rotate:
- img = img.rotate(degrees)
- # img.show()
- # img.save(os.path.join(save_paths[camn], all_images[camn][i]), compression='lzw')
- # img.close()
-
- # Stitch various cam views
- if prev_img is None:
- prev_img = img.copy()
-
- else:
- dst = Image.new('RGB', (prev_img.width + img.width, prev_img.height))
- dst.paste(prev_img, (0, 0))
- dst.paste(img, (prev_img.width, 0))
- prev_img = dst.copy()
-
- if not prev_img is None:
- save_name = all_images[self.main_camn][i][:-len(self.image_format) - 1] \
- + '-' + obj_name + '.{0}'.format(self.image_format)
- prev_img.save(os.path.join(save_paths[self.main_camn], save_name), compression='lzw')
- prev_img.close()
- img.close()
-
- def filter_interp_params(self, params, param_names, interpolate_nans, filter_high_freq):
- if not interpolate_nans and not filter_high_freq: return params
- new_params = copy.deepcopy(params)
-
- if self.show_plot: fig, axs = plt.subplots(len(param_names), 1)
-
- for i, label in enumerate(param_names):
- is_nan, x = np.isnan(new_params[label]), lambda z: z.nonzero()[0]
- if sum(~is_nan) == 0: continue # only nans
-
- if self.show_plot: axs[i].plot(new_params[label], label='raw')
-
- if '_a' in label:
- # fig2 = plt.figure()
- # ax2 = fig2.add_subplot(111)
- # ax2.plot(new_params[label], label='raw')
-
- new_params[label][~is_nan] = np.unwrap(np.deg2rad(new_params[label][~is_nan]))
- # ax2.plot(np.rad2deg(new_params[label]), label='unwrapped')
-
- if sum(is_nan) > 0: # interpolation needed for filter_high_freq
- new_params[label][is_nan] = np.interp(x(is_nan), x(~is_nan), new_params[label][~is_nan])
-
- if filter_high_freq and np.nanstd(new_params[label]) > 10e-6:
- w = self.cutoff_frequency / (self.framerate / 2) # Normalize the frequency
- b, a = signal.butter(2, w, 'low')
-
- new_params[label] = signal.filtfilt(b, a, new_params[label])
-
- if not interpolate_nans and sum(is_nan) > 0: # to fill back with nan if didnt want to interpolate
- new_params[label][is_nan] = np.nan
-
- if '_a' in label: # Wrap angles on [-pi, pi)
- new_params[label] = np.rad2deg((new_params[label] + np.pi) % (2 * np.pi) - np.pi)
- # ax2.plot(new_params[label], label='wrapped')
- # ax2.set_ylabel(label)
- # ax2.set_xlabel('frames')
-
- if self.show_plot:
- axs[i].plot(new_params[label], label='filt')
- axs[i].set_ylabel(label)
- axs[i].set_xlabel('frames')
-
- if self.show_plot:
- plt.legend()
- # plt.show()
-
- return new_params
-
- def save_skeleton_params_in_csv(self, body_params, wings_params, save_path):
- for obj_name in self.skeleton3d['obj_names']:
- body_params_names = body_params[obj_name].keys()
-
- params_list, params_names = [self.skeleton3d[obj_name]['frames']], 'frame'
- for param_name in body_params_names:
- params_list.append(body_params[obj_name][param_name])
- params_names += ',' + param_name
-
- for side in self.wings_params_init.keys():
- for param_name in wings_params[obj_name][side].keys():
- if param_name in body_params_names: continue
- params_list.append(wings_params[obj_name][side][param_name])
- params_names += ',' + param_name + '_' + side
-
- np.savetxt(os.path.join(save_path, 'skeleton_parameters-' + obj_name + '.csv'),
- np.c_[np.transpose(params_list)], delimiter=',', header=params_names)
-
- def save_obj_metric_in_csv(self, list_dict, list_name, save_path):
- for obj_name in self.skeleton3d['obj_names']:
- params_list, params_names = [self.skeleton3d[obj_name]['frames']], 'frame'
- if isinstance(list_dict[obj_name], dict):
- for side in self.wings_params_init.keys():
- params_list.append(list_dict[obj_name][side])
- params_names += ',' + list_name + '_' + side
- else:
- params_list.append(list_dict[obj_name])
- params_names += ',' + list_name
-
- np.savetxt(os.path.join(save_path, list_name + '-' + obj_name + '.csv'),
- np.c_[np.transpose(params_list)], delimiter=',', header=params_names)
-
- def stitch_images(self, images_names, image_paths, image_save_paths):
- ref_images, refs_images = set(), {}
- for camn in range(1, self.nb_cam + 1):
- refs_images[camn] = [image_name[20:-(len(self.image_format) + 1)] for image_name in images_names[camn]]
- ref_images = ref_images & set(refs_images[camn]) if len(ref_images) > 0 else set(refs_images[camn])
-
- ref_images = np.sort(list(ref_images))
-
- index_i = {}
- for camn in range(1, self.nb_cam + 1):
- index_i[camn] = [refs_images[camn].index(ref_image) for ref_image in ref_images]
-
- for i, _ in enumerate(index_i[self.main_camn]):
- imgs = {}
- for camn in range(1, self.nb_cam + 1):
- imgs[camn] = Image.open(os.path.join(image_paths[camn], images_names[camn][index_i[camn][i]]))
-
- img = img_utils.stitch_images(imgs, self.nb_cam)
- img.save(os.path.join(image_save_paths[self.main_camn], images_names[self.main_camn][index_i[self.main_camn][i]]), compression='lzw')
-
- def interp_2d_obj(self, rec_path, csv_path, save_path):
- rec_file_names = sorted(os.listdir(rec_path), key=lambda x: x[20:20+self.nb_leading_zero])
- rec_image_names = [s for s in rec_file_names if '.{0}'.format(self.image_format) in s]
- frames = [int(s[20:20+self.nb_leading_zero]) for s in rec_image_names]
-
- filenames_csv_path = os.listdir(csv_path)
- image_names_csv_path = [s for s in filenames_csv_path if '.{0}'.format(self.image_format) in s]
- image_names_csv_path = sorted(image_names_csv_path, key=lambda x: x[20:20+self.nb_leading_zero])
-
- csv_names = glob.glob(os.path.join(csv_path, '*-*-2d_points.csv'))
- csv_names = [csv_name[len(csv_path) + 1:] for csv_name in csv_names]
-
- obj_names = [csv_name[csv_name.find('-') + 1:csv_name.find('-2d_points.csv')] for csv_name in csv_names]
-
- coord_objs = {}
- if len(obj_names) == 0:
- for i, csv_name in enumerate(csv_names): coord_objs[obj_names[i]] = {}
- return coord_objs, obj_names
-
- img = Image.open(os.path.join(rec_path, rec_image_names[0]))
- if len(image_names_csv_path) == 0:
- resize_ratio = 1
- else:
- img_down = Image.open(os.path.join(csv_path, image_names_csv_path[0]))
- resize_ratio = img_down.size[0] / img.size[0]
-
- for i, csv_name in enumerate(csv_names):
- coord_objs[obj_names[i]] = \
- self.interpolate_coord(img, resize_ratio, frames, csv_name, csv_path, save_path)
-
- return coord_objs, obj_names
-
- def interpolate_coord(self, img, resize_ratio, frames, csv_name, csv_path, save_path):
- width_img, height_img = img.size
-
- old_obj_dict = np.genfromtxt(os.path.join(csv_path, csv_name), delimiter=',', skip_header=0, names=True)
-
- # TODO use resize_ratio in interpolation
- x_px = np.interp(frames, old_obj_dict['frame'], old_obj_dict['x_px'])
- y_px = np.interp(frames, old_obj_dict['frame'], old_obj_dict['y_px'])
-
- for i, _ in enumerate(frames):
- if x_px[i] > width_img: x_px[i] = width_img
- if y_px[i] > height_img: y_px[i] = height_img
- if x_px[i] < 0: x_px[i] = 0
- if y_px[i] < 0: y_px[i] = 0
-
- np.savetxt(os.path.join(save_path, csv_name), np.c_[frames, x_px, y_px], delimiter=',', header='frame,x_px,y_px')
- obj_dict = {'frame': frames, 'x_px': x_px, 'y_px': y_px}
-
- if self.show_plot:
- fig = plt.figure()
- ax = fig.add_subplot(111)
- plt.imshow(img)
- ax.scatter(obj_dict['x_px'], obj_dict['y_px'], marker='*')
- ax.scatter(old_obj_dict['x_px'], old_obj_dict['y_px'], marker='x')
- ax.set_xlabel('x (m)')
- ax.set_ylabel('y (m)')
- plt.xlim(0, width_img)
- plt.ylim(height_img, 0)
- plt.show()
-
- return obj_dict
-
- def nearest_date(self, dates_dict, date):
- diff_s = [abs(time.mktime(dates_dict[i]) - time.mktime(date)) for i in dates_dict]
- near_diff_s = [x for x in diff_s if x < self.max_diff_date_s]
-
- if len(near_diff_s) is 0:
- return []
- else:
- return dates_dict[diff_s.index(min(near_diff_s))]
-
-
-if __name__ == '__main__':
-
- img_process = ImagesProcessing()
-
- #path = 'G:'
- #img_process.cam_paths = {1: os.path.join(path, 'test'), 2: os.path.join(path, 'test'), 3: os.path.join(path, 'test')}
- #img_process.cam_save_paths = {1: os.path.join(path, 'test', '_Process'), 2: os.path.join(path, 'test', '_Process'), 3: os.path.join(path, 'test', '_Process')}
- #img_process.cam_paths = {1: 'F:\\Photron1', 2: 'Y:\\', 3: 'Z:\\'}
- #img_process.cam_save_paths = {1: 'F:\\Photron1\\_Process', 2: 'Y:\\_Process', 3: 'Z:\\_Process'}
- img_process.cam_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1',
- 2: '/media/user/MosquitoEscape_Photron2/Photron2',
- 3: '/media/user/MosquitoEscape_Photron3/Photron3'}
- img_process.cam_save_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1/_Process',
- 2: '/media/user/MosquitoEscape_Photron2/Photron2/_Process',
- 3: '/media/user/MosquitoEscape_Photron3/Photron3/_Process'}
-
- #img_process.cam_paths = {1: 'F:\\Photron1'}
- #img_process.cam_save_paths = {1: 'F:\\Photron1\\_Process'}
-
- img_process.init_paths_names()
- rec_names_to_process = {}
- for camn in range(1, img_process.nb_cam + 1):
- rec_names_to_process[camn] = [s for s in img_process.all_folders_rec_cam[camn] if '20200303' in s]
-
- img_process.do_batch('resize', rec_names=rec_names_to_process, step_frame=100, resize_ratio=1)
- # img_process.do_batch('crop', rec_names=rec_names_to_process, height_crop=200, width_crop=200, old_step_frame=100)
- # img_process.do_batch('save_tiff', rec_names=rec_names_to_process)
-
- # img_process.do_batch('sample', rec_names=rec_names_to_process, step_frame=1)
-
-
-
-
diff --git a/images/utils.py b/images/utils.py
index 07534670ba7c71d3f855e7d0d9af4cb35d2477be..376d701c0e2dd68537572495caf0c13c443a340f 100644
--- a/images/utils.py
+++ b/images/utils.py
@@ -1,4 +1,6 @@
-import os
+import fnmatch
+from difflib import SequenceMatcher
+from typing import List, Dict
import numpy as np
from PIL import Image, ImageEnhance, ImageOps
@@ -66,8 +68,107 @@ def stitch_images(imgs, nb_cam):
return img
-def save_tiff(image_name, rec_path, save_path):
- img = Image.open(os.path.join(rec_path, image_name))
+def save_tiff(img, save_path):
array = np.uint8(np.array(img) / 256)
- Image.fromarray(array).save(os.path.join(save_path, image_name), compression='lzw')
+ Image.fromarray(array).save(save_path, compression='lzw')
+
+
+def find_common_substrings(str_list: List[str], keep_shifted_matched: bool = True, keep_only_counts_higher_than: int = None,
+ return_nun_matches_instead_occurrences: bool = False, ) -> Dict[str, int]:
+
+ substring_counts = {}
+ for i in range(0, len(str_list)):
+
+ for j in range(i+1, len(str_list)):
+
+ string1 = str_list[i]
+ string2 = str_list[j]
+
+ match = SequenceMatcher(None, string1, string2).find_longest_match(0, len(string1), 0, len(string2))
+
+ if not keep_shifted_matched and match.a != match.b: continue
+
+ matching_substring = string1[match.a:match.a+match.size]
+
+ if matching_substring not in substring_counts:
+ substring_counts[matching_substring] = 1
+ else:
+ substring_counts[matching_substring] += 1
+
+ if keep_only_counts_higher_than is not None:
+ substring_counts = {key: substring_counts[key] for key in substring_counts.keys()
+ if substring_counts[key] >= keep_only_counts_higher_than}
+
+ if not return_nun_matches_instead_occurrences:
+ for key in substring_counts.keys():
+ substring_counts[key] = sum(key in s for s in str_list)
+
+ # Sort in descending order
+ substring_counts = dict(sorted(substring_counts.items(), key=lambda item: item[1], reverse=True))
+
+ return substring_counts
+
+
+def find_common_substrings_with_one_separator(str_list: List[str], return_nun_matches_instead_occurrences: bool = False,
+ keep_only_counts_higher_than: int = None) -> Dict[str, int]:
+ """
+ Fill find the common substring with separator (*), so is searching for blocks of two longest
+ substring (non shifted) matches between all strings in the given list
+
+ Args:
+ str_list: List of strings, usually file names
+ (e.g. ['cam1_202205040001-obj1.tif', 'cam1_202205040002-obj1.tif', 'cam1_202205040003-obj1.tif', ...]
+ -> then most common substrings with separator is 'cam1_20220504000*-obj1.tif')
+ return_nun_matches_instead_occurence:
+ keep_only_counts_higher_than:
+ """
+
+ substring_counts = {}
+ for i in range(0, len(str_list)):
+
+ for j in range(i+1, len(str_list)):
+
+ string1 = str_list[i]
+ string2 = str_list[j]
+
+ matches = SequenceMatcher(None, string1, string2).get_matching_blocks()
+
+ if len(matches)-1 != 2: continue # If there is more or less than two common substrings
+ if matches[0].a != matches[0].b and matches[1].a != matches[1].b: continue # If the matches are shifted
+
+ matching_substring = string1[matches[0].a:matches[0].a+matches[0].size] + '*' + string1[matches[1].a:matches[1].a+matches[1].size]
+
+ if matching_substring not in substring_counts:
+ substring_counts[matching_substring] = 1
+ else:
+ substring_counts[matching_substring] += 1
+
+ if keep_only_counts_higher_than is not None:
+ substring_counts = {key: substring_counts[key] for key in substring_counts.keys()
+ if substring_counts[key] >= keep_only_counts_higher_than}
+
+ if not return_nun_matches_instead_occurrences:
+ for key in substring_counts.keys():
+ substring_counts[key] = len(fnmatch.filter(str_list, key))
+
+ # Sort in descending order
+ substring_counts = dict(sorted(substring_counts.items(), key=lambda item: item[1], reverse=True))
+
+ return substring_counts
+
+
+def find_uncommon_substrings(str_list: List[str], common_substring) -> List[str]:
+
+ ind_star = common_substring.find('*')
+ uncommon_substrings = [s[ind_star:ind_star-len(common_substring)+1] for s in str_list]
+
+ return uncommon_substrings
+
+
+def find_file_numbers(str_list: List[str], common_substring) -> List[int]:
+
+ ind_star = common_substring.find('*')
+ file_numbers = [int(s[ind_star:s.rfind('.')]) for s in str_list]
+
+ return file_numbers
diff --git a/main_process_mating.py b/main_process_mating.py
index 93f5e0b8c8a5c49e7c2b554d2de85c3e46b5b1d6..e719a1ff0e87ac25c839808de9df205533abb8ca 100644
--- a/main_process_mating.py
+++ b/main_process_mating.py
@@ -1,14 +1,16 @@
import os, time
-from images.process import ImagesProcessing
+from process.batch_processing import BatchProcessing
#import deeplabcut
+# TODO Update from examples in test_process_batch
+
start = time.time()
-xyz_path = os.path.join(os.getcwd(), 'data/calib/xyz_calibration_device_small.csv')
+xyz_path = os.path.join(os.getcwd(), 'data/calib/examples/mosquito_escapes/to_delete/xyz_calibration_device_small.csv')
#xy_path = os.path.join(os.getcwd(), 'data/calib/20200615_xypts-rot.csv')
-xy_path = os.path.join(os.getcwd(), 'data/calib/20200615_xypts.csv')
-dlt_path = os.path.join(os.getcwd(), 'data/calib/20200615_DLTcoefs-py.csv')
+xy_path = os.path.join(os.getcwd(), 'data/calib/examples/mosquito_escapes/to_delete/20200615_xypts.csv')
+dlt_path = os.path.join(os.getcwd(), 'data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs-py.csv')
dlc_cfg_path = '/home/user/Desktop/Antoine/_DLC/config.yaml'
img_size = (896, 896)
@@ -16,7 +18,7 @@ img_size = (896, 896)
# Generate dlt coefficients
#gen_dlt(3, img_size, xyz_path, xy_path, dlt_path)
-img_process = ImagesProcessing(3, dlt_path)
+img_process = BatchProcessing()
img_process.cam_paths = {1: '/media/user/MosquitoLanding_Photron3_Backup/Photron1/_MatingKinematics',
2: '/media/user/MosquitoLanding_Photron3_Backup/Photron2/_MatingKinematics',
3: '/media/user/MosquitoLanding_Photron3_Backup/Photron3/_MatingKinematics'}
@@ -59,26 +61,26 @@ for camn in range(1, img_process.nb_cam + 1):
# > do everything automatically use 'multi_processes'
# Do cropping, rotating, stitching, etc at once and save as .avi (save much time and space)
-img_process.do_batch('multi_processes', rec_names=rec_names_to_process, yaml_path='processes.yaml', delete_previous=True)
+img_process._do_batch('multi_processes', rec_names_to_process=rec_names_to_process, yaml_path='processes.yaml', delete_previous=True)
# > OR do each step separately
-# Convert images to 8 bits and reduce framerate (save in cam_save_paths/_Sample)
-# img_process.do_batch('sample', rec_names=rec_names_to_process, step_frame=50, delete_previous=True)
+# Convert images to 8 bits and reduce frame_rate (save in cam_save_paths/_Sample)
+# img_process.do_batch('sample', rec_names_to_process=rec_names_to_process, step_frame=50, delete_previous=True)
#
# # Do 2d tracking (blob detection) on images in cam_save_paths/_Sample
-# img_process.do_batch('track2d', rec_names=rec_names_to_process, from_fn_name='sample')
+# img_process.do_batch('track2d', rec_names_to_process=rec_names_to_process, from_fn_name='sample')
#
# # Do 3d reconstruction of tracks
-# img_process.do_batch('recon3d', rec_names=rec_names_to_process, from_fn_name='sample')
+# img_process.do_batch('recon3d', rec_names_to_process=rec_names_to_process, from_fn_name='sample', dlt_path=dlt_path)
#
# # Crop all frames and save in cam_save_paths/_Cropped
-# img_process.do_batch('crop', rec_names=rec_names_to_process, from_fn_name='sample', height_crop=200, width_crop=200, delete_previous=True)
+# img_process.do_batch('crop', rec_names_to_process=rec_names_to_process, from_fn_name='sample', height_crop=200, width_crop=200, delete_previous=True)
#
# # Rotate view 2 (to always have piston coming from right side)
-# img_process.do_batch('rotate', rec_names=rec_names_to_process, from_fn_name='crop', camns=[2], degrees=270)
+# img_process.do_batch('rotate', rec_names_to_process=rec_names_to_process, from_fn_name='crop', camn_to_process=[2], degrees=270)
#
# # Stitch all views together and save in cam_save_paths/_Stitched
-# img_process.do_batch('stitch', rec_names=rec_names_to_process, from_fn_name='crop', delete_previous=True)
+# img_process.do_batch('stitch', rec_names_to_process=rec_names_to_process, from_fn_name='crop', delete_previous=True)
# -----------------------------------------------------------------------------------------------------------------------
@@ -97,18 +99,18 @@ wing_param_names = ['stroke_a', 'deviation_a', 'rotation_a']
# # Track features using DeepLabCut
-# img_process.do_batch('analyse_dlc', rec_names=rec_names_to_process, from_fn_name='save_avi', cfg_path=dlc_cfg_path,
-# shuffle=shuffle, trainingsetindex=0, batchsize=5, save_avi=False, model_name=model_name, delete_previous=True)
+# img_process.do_batch('analyse_dlc', rec_names_to_process=rec_names_to_process, from_fn_name='save_avi', cfg_path=dlc_cfg_path,
+# shuffle=shuffle, trainingset_index=0, batch_size=5, save_avi=False, model_name=model_name, delete_previous=True)
# # Load (+ filtering and unscrambling) 2d coords from DLC + Reverse processes (unstitch, rotate back, uncrop) + Reconstruct 3d coord
-# img_process.do_batch('load_dlc', rec_names=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
-# tracker_method='skeleton')
+# img_process.do_batch('load_dlc', rec_names_to_process=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
+# tracker_method='skeleton', dlt_path=dlt_path)
# # Optimize fit of skeleton to find body and wings angles
# img_process.multiprocessing = True
# img_process.threshold_likelihood = 0.85
-# img_process.do_batch('fit_skeleton', rec_names=rec_names_to_process, from_fn_name='load_dlc', model_name=model_name,
+# img_process.do_batch('fit_skeleton', rec_names_to_process=rec_names_to_process, from_fn_name='load_dlc', model_name=model_name,
# csv_path=csv_wings_geo_path, body_param_names=body_param_names, wing_param_names=wing_param_names,
-# animal_name="fly", res_method=res_method, opt_method=opt_method)
+# animal_name="fly", res_method=res_method, opt_method=opt_method, dlt_path=dlt_path)
print('> All processes as been processed (total time elapsed: {0:.4f} s)'.format(time.time() - start))
\ No newline at end of file
diff --git a/point_tracker/__init__.py b/point_tracker/__init__.py
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..1d8a7a8d303df66d9cf01d8eec0855d2e822aed0 100644
--- a/point_tracker/__init__.py
+++ b/point_tracker/__init__.py
@@ -0,0 +1,2 @@
+from point_tracker.tracker2d import Tracker2D
+from point_tracker.reconstructor3d import Reconstructor3D
diff --git a/point_tracker/reconstructor3d.py b/point_tracker/reconstructor3d.py
index 2887d0014ce3b16f6c633a02801ffb1c3a1c9d87..54c8dbe7062ec4e318727dce9cec770faa62aacf 100644
--- a/point_tracker/reconstructor3d.py
+++ b/point_tracker/reconstructor3d.py
@@ -1,69 +1,14 @@
-from __future__ import print_function
import os
+
import numpy as np
import matplotlib.pyplot as plt
+from mpl_toolkits.mplot3d import Axes3D
from dataclasses import dataclass
-from cam import calib
-
-
-def gen_dlt(nb_cam, img_size, xyz_path, xy_path, save_path, points_to_remove=[], plot_error_analysis=False):
- coord_calib_csv = np.genfromtxt(xyz_path, delimiter=',', skip_header=0, names=True)
- pts_calib_csv = np.genfromtxt(xy_path, delimiter=',', skip_header=0, names=True)
-
- coord_calib = {'x': np.array([s for i, s in enumerate(coord_calib_csv['x']) if i+1 not in points_to_remove]),
- 'y': np.array([s for i, s in enumerate(coord_calib_csv['y']) if i+1 not in points_to_remove]),
- 'z': np.array([s for i, s in enumerate(coord_calib_csv['z']) if i+1 not in points_to_remove])}
-
- reconstructor = Reconstructor3D(nb_cam, img_size)
-
- if plot_error_analysis: errors = {}
-
- pts_calib = {}
- for camn in range(1, nb_cam + 1):
- pts_calib['cam{0}'.format(camn)] = \
- {'x': np.array([s for i, s in enumerate(pts_calib_csv['cam{0}_X'.format(camn)]) if i+1 not in points_to_remove]),
- 'y': np.array([s for i, s in enumerate(pts_calib_csv['cam{0}_Y'.format(camn)]) if i+1 not in points_to_remove])}
-
- # Fill img_pts_calib with pixel coordinates from Matlab tracking (have to transform coordinate system)
- img_pts_calib = np.array([pts_calib['cam{0}'.format(camn)]['x'],
- img_size[1] - pts_calib['cam{0}'.format(camn)]['y']], np.float32).T
-
- # Remove nan
- index = ~np.isnan(img_pts_calib).any(axis=1)
- img_pts_calib = img_pts_calib[index]
-
- obj_coord_calib = np.array([coord_calib['x'][index], coord_calib['y'][index], coord_calib['z'][index]]).T
-
- _, rmse = reconstructor.gen_dltcoef(camn, obj_coord_calib, img_pts_calib)
-
- print('>> The reprojection error of cam{0} is {1} pixels'.format(camn, rmse))
+from camera import calib
+from camera.calib import read_dlt_coefs, gen_dlt_coefs_from_paths, save_dlt_coefs
- if plot_error_analysis:
- errors[camn] = np.zeros(len(img_pts_calib.T[0]))
- for i in range(len(img_pts_calib.T[0])):
- img_pts_calib_wo_pt = np.array([[s for j, s in enumerate(img_pts_calib.T[0]) if j is not i],
- [s for j, s in enumerate(img_pts_calib.T[0]) if j is not i]]).T
- obj_coord_calib_wo_pt = np.array([[s for j, s in enumerate(obj_coord_calib.T[0]) if j is not i],
- [s for j, s in enumerate(obj_coord_calib.T[1]) if j is not i],
- [s for j, s in enumerate(obj_coord_calib.T[2]) if j is not i]]).T
-
- _, errors[camn][i] = reconstructor.gen_dltcoef(camn, obj_coord_calib_wo_pt, img_pts_calib_wo_pt)
-
- reconstructor.save_dltcoef(save_path)
- print('> DLT coefficient have been saved in {0}'.format(save_path))
-
- if plot_error_analysis:
- # Plot all rmse without one point
-
- for camn in range(1, nb_cam + 1):
- fig = plt.figure('Error analysis (cam{0})'.format(camn))
- ax = fig.add_subplot(111)
- ax.scatter(list(range(1, len(errors[camn]) +1)), errors[camn])
- ax.set_xlabel('x (m)')
- ax.set_ylabel('y (m)')
- plt.show()
@dataclass
class Reconstructor3DSettings:
@@ -72,28 +17,39 @@ class Reconstructor3DSettings:
threshold_delay_tracks: int = 10 # max delay (in frames) between two successive 3d points to set as a track
threshold_length_tracks: int = 5
+ image_ratio: float = 1.0 # ratio between size of the images used for calibration over the analyzed images (if changed).
+
+ @staticmethod
+ def from_dict(dictionary):
+ reconstructor_sets = Reconstructor3DSettings()
+ for key, value in dictionary.items():
+ setattr(reconstructor_sets, key, value)
+
+ return reconstructor_sets
+
class Reconstructor3D:
- def __init__(self, nb_cam, img_size):
+ def __init__(self, dlt_coefs):
"""
Class used to reconstruct 3d coordinates of object from 2d coordinates on images
Args:
- nb_cam: number of camera (minimum 2)
- img_size: size of the images in pixel (width, height)
+ dlt_coefs: DLT coefficients (11*nb_cam)
"""
- self.nb_cam = nb_cam
- self.img_size = img_size
+ assert len(dlt_coefs[0]) == 11
+
+ self.dlt_coefs = dlt_coefs
+ self.nb_cam = len(self.dlt_coefs)
# TODO this use of dataclass Reconstructor3DSettings doesn't seems really useful
- self.recon3d_settings = Reconstructor3DSettings()
+ self.settings = Reconstructor3DSettings()
- self.dlt_coefs = np.zeros((self.nb_cam, 12))
self.pts_dict = {}
self.objs_dict = {}
- def read_dltcoef(self, path):
+ @staticmethod
+ def read_dlt_coefs(path) -> "Reconstructor3D":
"""
Read DLT coefficients from .csv file
@@ -101,49 +57,9 @@ class Reconstructor3D:
path: path of the .csv file with DLT (11) coefficients (along rows) for all camera (along column)
"""
- dlt_csv = np.genfromtxt(path, delimiter=',', skip_header=0, names=True)
- dlt_csv.dtype.names = ['cam{0}'.format(camn) for camn in range(1, self.nb_cam + 1)]
-
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- self.dlt_coefs[i] = np.append(np.array(dlt_csv[camn]), 1)
-
- def save_dltcoef(self, path):
- """
- Save DLT coefficients in .csv file
-
- Args:
- path: path of the .csv file with DLT (11) coefficients (along rows) for all camera (along column)
- """
-
- dlt_coefs = np.zeros((self.nb_cam, 11))
- for i, camn in enumerate(range(1, self.nb_cam + 1)):
- dlt_coefs[i] = self.dlt_coefs[:-1]
-
- header = ['cam{0}'.format(camn) for camn in range(1, self.nb_cam + 1)]
- header = ','.join(header)
-
- np.savetxt(path, np.c_[dlt_coefs].T, delimiter=',', header=header)
-
- def gen_dltcoef(self, camn, obj_coord, img_pts):
- """
- Generate DLT coefficients for one camera
-
- args are vectors of vectors (e.g. np.array([[451, 123], [78, 45], [78, 79]])) of size nb_obj*nb_coord
-
- Args:
- camn: num of camera
- obj_coord: known 3d coordinates of the objects
- img_pts: 2d coordinates for all objects for one camera
-
- Returns:
- dlt_coefs_cam: DLT coefficient (12: 11 + 1(= 1.0)) for one camera
- mean_repro_err: mean reprojcetion error in pixel (or Root-mean-square deviation)
- """
-
- dlt_coefs_cam, repro_err = calib.calib(3, obj_coord, img_pts)
- self.dlt_coefs[camn-1] = dlt_coefs_cam
+ dlt_coefs = read_dlt_coefs(path)
- return dlt_coefs_cam, repro_err
+ return Reconstructor3D(dlt_coefs)
def recon_objs(self, xs_pts, ys_pts, frames_coord, with_missing=True, sort_by_rmse=True):
"""
@@ -271,8 +187,8 @@ class Reconstructor3D:
#xs_px_obj, ys_px_obj = [], []
xs_obj, ys_obj, zs_obj, rmses_obj, index_match_pts = [], [], [], [], np.array([])
- for num_combi in range(0, len(combi_indexes_frames)):
- index_match = list(combi_indexes_frames[num_combi])
+ for nb_combi in range(0, len(combi_indexes_frames)):
+ index_match = list(combi_indexes_frames[nb_combi])
index_match2 = index_match.copy()
for i in range(0, self.nb_cam):
@@ -293,9 +209,11 @@ class Reconstructor3D:
pts_cam = np.array([xs_pts[i][index_match[i]], ys_pts[i][index_match[i]]]).T
pts_cams = np.vstack([pts_cams, pts_cam]) if pts_cams.size > 0 else pts_cam
- obj_coord = calib.recon3D(3, self.nb_cam - count_missing, cur_Ls, pts_cams)
+ obj_coord = calib.recon3d(3, self.nb_cam - count_missing, cur_Ls, pts_cams*self.settings.image_ratio)
repro_pts_cams = calib.find2d(self.nb_cam, self.dlt_coefs, np.array([obj_coord]))
+ repro_pts_cams = repro_pts_cams/self.settings.image_ratio
+
repro_pts_cams_wo_missing = \
np.array([repro_pts_cams[j] for j, _ in enumerate(range(1, self.nb_cam + 1)) if index_match2[j] != None])
@@ -303,7 +221,7 @@ class Reconstructor3D:
correction_factor = count_missing * np.sqrt(2) if count_missing > 0 else 1
mean_repro_err = np.sqrt(np.mean(np.sum((repro_pts_cams_wo_missing - pts_cams) ** 2, 1))) * correction_factor
- if mean_repro_err < self.recon3d_settings.threshold_mean_repro_err:
+ if mean_repro_err < self.settings.threshold_mean_repro_err:
# Fill tracking results
xs_obj.append(obj_coord[0]), ys_obj.append(obj_coord[1]), zs_obj.append(obj_coord[2])
rmses_obj.append(mean_repro_err)
@@ -327,7 +245,7 @@ class Reconstructor3D:
indexes_pts: index of the matched 2d pts used to reconstruct 3d coordinates for all frames
Returns:
- tracks_dict: dictionary with tracks of all objects (e.g. tracks_dict[num_obj] = {'x': [], ...})
+ tracks_dict: dictionary with tracks of all objects (e.g. tracks_dict[nb_obj] = {'x': [], ...})
"""
self.tracks_dict = {'obj': {}, 'obj_names': []}
@@ -341,15 +259,15 @@ class Reconstructor3D:
self.tracks_dict = {'obj': {}, 'obj_names': []}
return self.tracks_dict
- num_objs_live = []
+ nb_objs_live = []
min_frame_diff = np.min(np.array(frames_objs[1:]) - np.array(frames_objs[:-1]))
for i, frame in enumerate(frames_objs):
- if len(num_objs_live) == 0: # to initialize first track
+ if len(nb_objs_live) == 0: # to initialize first track
for j in range(0, len(xs_objs[i])):
- num_objs_live.append(j + 1)
- self.tracks_dict['obj'][num_objs_live[-1]] = {'x': [xs_objs[i][j]], 'y': [ys_objs[i][j]],
+ nb_objs_live.append(j + 1)
+ self.tracks_dict['obj'][nb_objs_live[-1]] = {'x': [xs_objs[i][j]], 'y': [ys_objs[i][j]],
'z': [zs_objs[i][j]], 'frame': [frame], 'index_pts': [indexes_pts[i][j]]}
continue
@@ -360,66 +278,66 @@ class Reconstructor3D:
# then find best match for next obj with all points except the previously matched point
for j in range(0, len(xs_objs[i])):
distances = []
- for num_obj_live in num_objs_live:
- if self.tracks_dict['obj'][num_obj_live]['frame'][-1] == frame: # to avoid using same frame twice
+ for nb_obj_live in nb_objs_live:
+ if self.tracks_dict['obj'][nb_obj_live]['frame'][-1] == frame: # to avoid using same frame twice
distances.append(np.nan)
else:
- distances.append(np.sqrt((self.tracks_dict['obj'][num_obj_live]['x'][-1] - xs_objs[i][j]) ** 2
- + (self.tracks_dict['obj'][num_obj_live]['y'][-1] - ys_objs[i][j]) ** 2
- + (self.tracks_dict['obj'][num_obj_live]['z'][-1] - zs_objs[i][j]) ** 2))
+ distances.append(np.sqrt((self.tracks_dict['obj'][nb_obj_live]['x'][-1] - xs_objs[i][j]) ** 2
+ + (self.tracks_dict['obj'][nb_obj_live]['y'][-1] - ys_objs[i][j]) ** 2
+ + (self.tracks_dict['obj'][nb_obj_live]['z'][-1] - zs_objs[i][j]) ** 2))
# if distance between two successive object is small, then they are stitched together in one track
- if not np.isnan(distances).all() and np.nanmin(distances) < self.recon3d_settings.threshold_distance_tracks:
- num_obj_distance_min = num_objs_live[np.nanargmin(distances)]
+ if not np.isnan(distances).all() and np.nanmin(distances) < self.settings.threshold_distance_tracks:
+ nb_obj_distance_min = nb_objs_live[np.nanargmin(distances)]
- self.tracks_dict['obj'][num_obj_distance_min]['x'].append(xs_objs[i][j])
- self.tracks_dict['obj'][num_obj_distance_min]['y'].append(ys_objs[i][j])
- self.tracks_dict['obj'][num_obj_distance_min]['z'].append(zs_objs[i][j])
- self.tracks_dict['obj'][num_obj_distance_min]['frame'].append(frame)
- self.tracks_dict['obj'][num_obj_distance_min]['index_pts'] = \
- np.vstack([self.tracks_dict['obj'][num_obj_distance_min]['index_pts'], indexes_pts[i][j]])
+ self.tracks_dict['obj'][nb_obj_distance_min]['x'].append(xs_objs[i][j])
+ self.tracks_dict['obj'][nb_obj_distance_min]['y'].append(ys_objs[i][j])
+ self.tracks_dict['obj'][nb_obj_distance_min]['z'].append(zs_objs[i][j])
+ self.tracks_dict['obj'][nb_obj_distance_min]['frame'].append(frame)
+ self.tracks_dict['obj'][nb_obj_distance_min]['index_pts'] = \
+ np.vstack([self.tracks_dict['obj'][nb_obj_distance_min]['index_pts'], indexes_pts[i][j]])
else: # start a new tracks
- num_objs_live.append(num_objs_live[-1] + 1)
- self.tracks_dict['obj'][num_objs_live[-1]] = {'x': [xs_objs[i][j]], 'y': [ys_objs[i][j]],
+ nb_objs_live.append(nb_objs_live[-1] + 1)
+ self.tracks_dict['obj'][nb_objs_live[-1]] = {'x': [xs_objs[i][j]], 'y': [ys_objs[i][j]],
'z': [zs_objs[i][j]], 'frame': [frame], 'index_pts': [indexes_pts[i][j]]}
- # remove obj from num_objs_live if too old (more than a given number of frames)
- for k, num_obj_live in enumerate(num_objs_live):
- frame_diff = frame - self.tracks_dict['obj'][num_obj_live]['frame'][-1]
- if frame_diff > (self.recon3d_settings.threshold_delay_tracks * min_frame_diff):
- del num_objs_live[k]
+ # remove obj from nb_objs_live if too old (more than a given number of frames)
+ for k, nb_obj_live in enumerate(nb_objs_live):
+ frame_diff = frame - self.tracks_dict['obj'][nb_obj_live]['frame'][-1]
+ if frame_diff > (self.settings.threshold_delay_tracks * min_frame_diff):
+ del nb_objs_live[k]
# Get rid of tracks that are too short and add nan to other
- for num_obj in list(self.tracks_dict['obj'].keys()):
- if len(self.tracks_dict['obj'][num_obj]['x']) <= self.recon3d_settings.threshold_length_tracks:
- del self.tracks_dict['obj'][num_obj]
+ for nb_obj in list(self.tracks_dict['obj'].keys()):
+ if len(self.tracks_dict['obj'][nb_obj]['x']) <= self.settings.threshold_length_tracks:
+ del self.tracks_dict['obj'][nb_obj]
else:
- self.tracks_dict['obj_names'].append('obj{0}'.format(num_obj))
+ self.tracks_dict['obj_names'].append('obj{0}'.format(nb_obj))
- frames_obj = self.tracks_dict['obj'][num_obj]['frame']
+ frames_obj = self.tracks_dict['obj'][nb_obj]['frame']
frames_obj = np.arange(np.min(frames_obj), np.max(frames_obj), min_frame_diff)
xs_objs, ys_objs, zs_objs, index_pts = [], [], [], []
for i, frame in enumerate(frames_obj):
- in_frame = np.squeeze(np.where(frame == self.tracks_dict['obj'][num_obj]['frame']))
+ in_frame = np.squeeze(np.where(frame == self.tracks_dict['obj'][nb_obj]['frame']))
if in_frame.shape == (0,):
xs_objs.append(np.nan), ys_objs.append(np.nan), zs_objs.append(np.nan)
index_pts.append([None, None, None])
else:
- xs_objs.append(self.tracks_dict['obj'][num_obj]['x'][in_frame])
- ys_objs.append(self.tracks_dict['obj'][num_obj]['y'][in_frame])
- zs_objs.append(self.tracks_dict['obj'][num_obj]['z'][in_frame])
- index_pts.append(self.tracks_dict['obj'][num_obj]['index_pts'][in_frame])
+ xs_objs.append(self.tracks_dict['obj'][nb_obj]['x'][in_frame])
+ ys_objs.append(self.tracks_dict['obj'][nb_obj]['y'][in_frame])
+ zs_objs.append(self.tracks_dict['obj'][nb_obj]['z'][in_frame])
+ index_pts.append(self.tracks_dict['obj'][nb_obj]['index_pts'][in_frame])
- self.tracks_dict['obj'][num_obj]['x'] = xs_objs
- self.tracks_dict['obj'][num_obj]['y'] = ys_objs
- self.tracks_dict['obj'][num_obj]['z'] = zs_objs
- self.tracks_dict['obj'][num_obj]['frame'] = frames_obj
- self.tracks_dict['obj'][num_obj]['index_pts'] = index_pts
+ self.tracks_dict['obj'][nb_obj]['x'] = xs_objs
+ self.tracks_dict['obj'][nb_obj]['y'] = ys_objs
+ self.tracks_dict['obj'][nb_obj]['z'] = zs_objs
+ self.tracks_dict['obj'][nb_obj]['frame'] = frames_obj
+ self.tracks_dict['obj'][nb_obj]['index_pts'] = index_pts
print(">>> {0} track(s) have been kept (out of {1} reconstructed tracks)"
- .format(len(self.tracks_dict['obj_names']), num_objs_live[-1]))
+ .format(len(self.tracks_dict['obj_names']), nb_objs_live[-1]))
return self.tracks_dict
@@ -445,10 +363,13 @@ class Reconstructor3D:
for j, camn in enumerate(range(1, self.nb_cam + 1)):
repro_pts_cam = calib.find2d(1, self.dlt_coefs[j], obj_coord).T
+ repro_pts_cams = repro_pts_cams/self.settings.image_ratio
xs_pts[j].append(repro_pts_cam[0][0])
ys_pts[j].append(repro_pts_cam[1][0])
+ xs_pts, ys_pts = xs_pts/self.settings.image_ratio, ys_pts/self.settings.image_ratio
+
return np.array(xs_pts), np.array(ys_pts)
def save_points_csv(self, names, paths, save_type='all_points'):
@@ -497,75 +418,47 @@ class Reconstructor3D:
paths: paths where to save .csv files
"""
- for num_obj in list(self.tracks_dict['obj'].keys()):
+ for nb_obj in list(self.tracks_dict['obj'].keys()):
obj_coords = np.array([])
- for i in range(0, len(self.tracks_dict['obj'][num_obj]['x'])):
- obj_coord = np.array([self.tracks_dict['obj'][num_obj]['x'][i],
- self.tracks_dict['obj'][num_obj]['y'][i],
- self.tracks_dict['obj'][num_obj]['z'][i]])
+ for i in range(0, len(self.tracks_dict['obj'][nb_obj]['x'])):
+ obj_coord = np.array([self.tracks_dict['obj'][nb_obj]['x'][i],
+ self.tracks_dict['obj'][nb_obj]['y'][i],
+ self.tracks_dict['obj'][nb_obj]['z'][i]])
obj_coords = np.vstack([obj_coords, obj_coord]) if obj_coords.size else obj_coord
for i, camn in enumerate(range(1, self.nb_cam + 1)):
repro_pts_cams = calib.find2d(1, self.dlt_coefs[i], np.array(obj_coords)).T
+ repro_pts_cams = repro_pts_cams/self.settings.image_ratio
- if not np.array_equal(self.tracks_dict['obj'][num_obj]['frame'],
- np.sort(np.unique(self.tracks_dict['obj'][num_obj]['frame']))):
+ if not np.array_equal(self.tracks_dict['obj'][nb_obj]['frame'],
+ np.sort(np.unique(self.tracks_dict['obj'][nb_obj]['frame']))):
raise ValueError('frames_objs must be sorted and unique')
- np.savetxt(os.path.join(paths[i], names[i] + '-obj{0}-2d_points.csv'.format(num_obj)),
- np.c_[self.tracks_dict['obj'][num_obj]['frame'], repro_pts_cams[0], repro_pts_cams[1]],
+ np.savetxt(os.path.join(paths[i], names[i] + '-obj{0}-2d_points.csv'.format(nb_obj)),
+ np.c_[self.tracks_dict['obj'][nb_obj]['frame'], repro_pts_cams[0], repro_pts_cams[1]],
delimiter=',', header='frame,x_px,y_px')
- np.savetxt(os.path.join(paths[i], names[i] + '-obj{0}-3d_track.csv'.format(num_obj)),
- np.c_[self.tracks_dict['obj'][num_obj]['frame'], self.tracks_dict['obj'][num_obj]['x'],
- self.tracks_dict['obj'][num_obj]['y'], self.tracks_dict['obj'][num_obj]['z']],
+ np.savetxt(os.path.join(paths[i], names[i] + '-obj{0}-3d_track.csv'.format(nb_obj)),
+ np.c_[self.tracks_dict['obj'][nb_obj]['frame'], self.tracks_dict['obj'][nb_obj]['x'],
+ self.tracks_dict['obj'][nb_obj]['y'], self.tracks_dict['obj'][nb_obj]['z']],
delimiter=',', header='frame,x,y,z')
if __name__ == '__main__':
- image_format = 'tif'
- # rec_names = {1: 'cam1_20200303_030117'}
- # image_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1/_Process/_DownSized/'}
+ path = '../data/examples/mosquito_escapes/'
+ image_paths = {1: os.path.join(path, 'cam1'), 2: os.path.join(path, 'cam2'), 3: os.path.join(path, 'cam3')}
rec_names = {1: 'cam1_20200303_030117', 2: 'cam2_20200303_030120', 3: 'cam3_20200303_030120'}
- image_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1/_Process/_Sample/',
- 2: '/media/user/MosquitoEscape_Photron2/Photron2/_Process/_Sample/',
- 3: '/media/user/MosquitoEscape_Photron3/Photron3/_Process/_Sample/'}
-
- nb_cam = 3
- img_size = (1024, 1024)
-
- # coord_calib_csv = np.genfromtxt('data/calib/xyz_calibration_device_small.csv', delimiter=',', skip_header=0, names=True)
- # pts_calib_csv = np.genfromtxt('data/calib/20200401_xypts.csv', delimiter=',', skip_header=0, names=True)
- coord_calib_csv = np.genfromtxt('data/calib/xyz_calibration_device_big.csv', delimiter=',', skip_header=0, names=True)
- pts_calib_csv = np.genfromtxt('data/calib/20200305_xypts.csv', delimiter=',', skip_header=0, names=True)
-
- coord_calib = {'x': np.array(coord_calib_csv['x']), 'y': np.array(coord_calib_csv['y']),
- 'z': np.array(coord_calib_csv['z'])}
-
- reconstructor = Reconstructor3D(nb_cam, img_size)
-
- # generate calibration (DLT coefficients)
- pts_calib = {}
- for camn in range(1, nb_cam + 1):
- pts_calib[camn] = {'x': np.array(pts_calib_csv['cam{0}_X'.format(camn)]),
- 'y': np.array(pts_calib_csv['cam{0}_Y'.format(camn)])}
-
- # Fill img_pts_calib with pixel coordinates from Matlab tracking (have to transform coordinate system)
- img_pts_calib = np.array([pts_calib[camn]['x'], img_size[1] - pts_calib[camn]['y']], np.float32).T
-
- # Remove nan
- index = ~np.isnan(img_pts_calib).any(axis=1)
- img_pts_calib = img_pts_calib[index]
- obj_coord_calib = np.array([coord_calib['x'][index], coord_calib['y'][index], coord_calib['z'][index]]).T
+ coord_calib_csv_path = '../data/calib/examples/mosquito_escapes/xyz_calibration_device.csv'
+ pts_calib_csv_path = '../data/calib/examples/mosquito_escapes/20200306_xypts.csv'
- _, rmse = reconstructor.gen_dltcoef(camn, obj_coord_calib, img_pts_calib)
+ dlt_coefs, _ = gen_dlt_coefs_from_paths(coord_calib_csv_path, pts_calib_csv_path, from_matlab=True, img_height=1024)
+ save_dlt_coefs(dlt_coefs, '../data/calib/examples/mosquito_escapes/20200306_DLTcoefs-py.csv')
- print('>>> The mean reprojection error of cam{0} is {1:.2f} pixels'.format(camn, rmse))
+ nb_cam = len(rec_names.keys())
- #reconstructor.read_dltcoef('data/calib/20200305_DLTcoefs.csv')
- reconstructor.save_dltcoef('data/calib/20200305_DLTcoefs-py.csv')
+ reconstructor = Reconstructor3D(dlt_coefs)
# From .csv file, fill xs_pose, ys_pose and frames_pose (2d coordinates of blobs)
pts_csv = {}
@@ -582,13 +475,14 @@ if __name__ == '__main__':
tracks_dict = reconstructor.recon_tracks(xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts)
save_names = [rec_names[camn] for camn in range(1, nb_cam + 1)]
- save_paths = [image_paths[camn] for camn in range(1, nb_cam + 1)]
+ save_paths = [os.path.join(image_paths[camn], rec_names[camn]) for camn in range(1, nb_cam + 1)]
reconstructor.save_tracks_csv(save_names, save_paths)
fig = plt.figure()
- ax = fig.add_subplot(111, projection='3d')
- for num_obj in range(1, tracks_dict['nb_tracks'] + 1):
- ax.scatter(tracks_dict['obj'][num_obj]['x'], tracks_dict['obj'][num_obj]['y'], tracks_dict['obj'][num_obj]['z'])
+ #ax = fig.add_subplot(111, projection='3d')
+ ax = Axes3D(fig)
+ for obj_num in tracks_dict['obj'].keys():
+ ax.scatter(tracks_dict['obj'][obj_num]['x'], tracks_dict['obj'][obj_num]['y'], tracks_dict['obj'][obj_num]['z'])
# ax.scatter(xs_pose, ys_pose, zs_pose)
ax.set_xlabel('x (m)')
ax.set_ylabel('y (m)')
diff --git a/point_tracker/tracker2d.py b/point_tracker/tracker2d.py
index ce7e0e3e54f85e7916210e9ccc7b9de70b2f876a..6fe781b40725a227b44f557de0dcf3ec9ab536e7 100644
--- a/point_tracker/tracker2d.py
+++ b/point_tracker/tracker2d.py
@@ -1,20 +1,29 @@
-from __future__ import print_function
import os
-import cv2, glob
+import cv2
import numpy as np
from PIL import Image
from collections import defaultdict
+from typing import List
+
from dataclasses import dataclass
@dataclass
-class BackgroundSubstractorSettings:
+class BackgroundSubtractorSettings:
type: str = 'KNN' # 'KNN' or 'MOG2'
varThreshold: int = 12
dist2Threshold: int = 50
shadow_threshold: int = 1
+ @staticmethod
+ def from_dict(dictionary):
+ background_sets = BackgroundSubtractorSettings()
+ for key, value in dictionary.items():
+ setattr(background_sets, key, value)
+
+ return background_sets
+
@dataclass
class BlobDetectorSettings:
@@ -30,6 +39,14 @@ class BlobDetectorSettings:
aspect_ratio_min: float = 0.8
eucli_dist_max: float = 5.0
+ @staticmethod
+ def from_dict(dictionary):
+ blob_detector_sets = BlobDetectorSettings()
+ for key, value in dictionary.items():
+ setattr(blob_detector_sets, key, value)
+
+ return blob_detector_sets
+
class Tracker2D:
def __init__(self):
@@ -43,61 +60,48 @@ class Tracker2D:
self.points = defaultdict(list)
- self.rec_name = 'test'
- self.save_path = os.getcwd()
+ self.get_blob_detector(BlobDetectorSettings())
+ self.get_back_subtractor(BackgroundSubtractorSettings())
- self.back_substractor_settings = BackgroundSubstractorSettings()
- self.blob_detector_settings = BlobDetectorSettings()
-
- self.initialize()
-
- def initialize(self):
- """
-
- """
-
- self.get_blob_detector()
- self.get_back_substractor()
-
- def load_images(self, path, image_format):
+ def load_images(self, image_names: List[str], path: str, frames: List[int] = None) -> None:
"""
Args:
- path:
- image_format:
+ image_names: a List of image names
+ path: The path where the images are
+ frames: a list of int with the frame number corresponding to the images
"""
- image_paths = glob.glob(os.path.join(path, '*.{0}'.format(image_format)))
- image_names = [os.path.basename(image_path) for image_path in image_paths]
- self.rec_name = image_names[0][:20]
- self.save_path = path
+ self.images = [cv2.imread(os.path.join(path, image_name), cv2.IMREAD_GRAYSCALE) for image_name in image_names]
- self.images = [cv2.imread(image_path, cv2.IMREAD_GRAYSCALE) for image_path in image_paths]
- self.frames = [int(image_name[20:-len('.{0}'.format(image_format))]) for image_name in image_names]
+ if frames is None:
+ self.frames = list(range(1, len(self.images) +1))
+ else:
+ self.frames = frames
in_sort = np.argsort(self.frames)
self.images = [self.images[in_s] for in_s in in_sort]
self.frames = [self.frames[in_s] for in_s in in_sort]
- def do_tracking(self):
+ def do_tracking(self, ind_img_for_background: List[int] = None):
"""
"""
- if len(self.images) is 0:
- print("WARN: you need to load images first")
- return
+ assert len(self.images) != 0, "Images need to be loaded first"
+
+ if ind_img_for_background is None: ind_img_for_background = range(0, len(self.images))
- [self.back_substractor.apply(image, learningRate=-1) for image in self.images[:-6]]
+ [self.back_subtractor.apply(self.images[i], learningRate=-1) for i in ind_img_for_background]
- self.background = self.back_substractor.getBackgroundImage()
+ self.background = self.back_subtractor.getBackgroundImage()
self.background = cv2.medianBlur(self.background, 5) # Add median filter to images
self.points = defaultdict(list)
self.foreground_masks = self.images.copy()
for i, image in enumerate(self.images):
- self.foreground_masks[i] = self.back_substractor.apply(image, learningRate=-1)
+ self.foreground_masks[i] = self.back_subtractor.apply(image, learningRate=-1)
self.foreground_masks[i] = cv2.medianBlur(self.foreground_masks[i], 3) # Add median filter to images
blobs = self.blob_detector.detect((255 - self.foreground_masks[i])) # Detect blobs
@@ -131,14 +135,14 @@ class Tracker2D:
print("WARN: you need to track or load 2d points first")
return
- [self.back_substractor.apply(image, learningRate=-1) for image in self.images[:-6]]
+ [self.back_subtractor.apply(image, learningRate=-1) for image in self.images[:-6]]
if flip_time: # Will start analysing last frame
images, frames = np.flip(self.images, axis=0), np.flip(self.frames)
else:
images, frames = self.images, self.frames
- self.background = self.back_substractor.getBackgroundImage()
+ self.background = self.back_subtractor.getBackgroundImage()
self.background = cv2.medianBlur(self.background, 5) # Add median filter to images
if 'strip' in shape: height, width = images[0].shape
@@ -152,7 +156,7 @@ class Tracker2D:
for i, image in enumerate(images):
if not i % step == 0: continue
- self.foreground_masks[i] = self.back_substractor.apply(image, learningRate=0)
+ self.foreground_masks[i] = self.back_subtractor.apply(image, learningRate=0)
self.foreground_masks[i] = cv2.medianBlur(self.foreground_masks[i], 3) # Add median filter to images
in_frame = np.where(frames[i] == self.points['frame'])[0]
@@ -183,28 +187,28 @@ class Tracker2D:
self.strobe_img = self.strobe_img.astype(np.uint8)
- def get_back_substractor(self):
- if self.back_substractor_settings.type is 'MOG2':
- self.back_substractor = cv2.createBackgroundSubtractorMOG2(history=-1, varThreshold=self.back_substractor_settings.varThreshold, detectShadows=True)
+ def get_back_subtractor(self, back_subtractor_settings: BackgroundSubtractorSettings()):
+ if back_subtractor_settings.type is 'MOG2':
+ self.back_subtractor = cv2.createBackgroundSubtractorMOG2(history=-1, varThreshold=back_subtractor_settings.varThreshold, detectShadows=True)
- elif self.back_substractor_settings.type is 'KNN':
- self.back_substractor = cv2.createBackgroundSubtractorKNN(history=-1, dist2Threshold=self.back_substractor_settings.dist2Threshold, detectShadows=True)
- self.back_substractor.setShadowThreshold(self.back_substractor_settings.shadow_threshold) # 0.5 by default
+ elif back_subtractor_settings.type is 'KNN':
+ self.back_subtractor = cv2.createBackgroundSubtractorKNN(history=-1, dist2Threshold=back_subtractor_settings.dist2Threshold, detectShadows=True)
+ self.back_subtractor.setShadowThreshold(back_subtractor_settings.shadow_threshold) # 0.5 by default
- def get_blob_detector(self):
+ def get_blob_detector(self, blob_detector_settings: BlobDetectorSettings()):
# Setup SimpleBlobDetector parameters.
params = cv2.SimpleBlobDetector_Params()
- params.minThreshold = self.blob_detector_settings.threshold_min
- params.maxThreshold = self.blob_detector_settings.threshold_max
+ params.minThreshold = blob_detector_settings.threshold_min
+ params.maxThreshold = blob_detector_settings.threshold_max
params.filterByArea = True
- params.minArea = self.blob_detector_settings.area_min
- params.maxArea = self.blob_detector_settings.area_max
+ params.minArea = blob_detector_settings.area_min
+ params.maxArea = blob_detector_settings.area_max
params.filterByInertia = True
- params.minInertiaRatio = self.blob_detector_settings.inertia_ratio_min
- params.maxInertiaRatio = self.blob_detector_settings.inertia_ratio_max
+ params.minInertiaRatio = blob_detector_settings.inertia_ratio_min
+ params.maxInertiaRatio = blob_detector_settings.inertia_ratio_max
params.filterByCircularity = False
params.filterByConvexity = False
@@ -231,6 +235,7 @@ class Tracker2D:
if __name__ == '__main__':
+ import glob
image_format = 'tif'
# rec_name = 'cam1_20200303_030117'
@@ -242,11 +247,15 @@ if __name__ == '__main__':
3: '/media/user/MosquitoEscape_Photron3/Photron3/_Process/_DownSized/'}
tracker = Tracker2D()
- tracker.initialize()
for camn in range(1, nb_cam + 1):
print(os.path.join(image_paths[camn], rec_names[camn]))
- tracker.load_images(os.path.join(image_paths[camn], rec_names[camn]), image_format)
+
+ all_image_paths = glob.glob(os.path.join(image_paths[camn], '*.{0}'.format(image_format)))
+ image_names = [os.path.basename(image_path) for image_path in all_image_paths]
+ frames = [int(image_name[20:-len('.{0}'.format(image_format))]) for image_name in image_names]
+
+ tracker.load_images(image_names, image_paths[camn], frames)
tracker.do_tracking()
tracker.save_csv(rec_names[camn], os.path.join(image_paths[camn], rec_names[camn]))
diff --git a/analysis/__init__.py b/process/__init__.py
similarity index 100%
rename from analysis/__init__.py
rename to process/__init__.py
diff --git a/process/batch_processing.py b/process/batch_processing.py
new file mode 100644
index 0000000000000000000000000000000000000000..b8288d9e501c59f2cbbae788daec2bc1ca59d8bc
--- /dev/null
+++ b/process/batch_processing.py
@@ -0,0 +1,1090 @@
+import time, shutil, yaml, glob
+
+import cv2
+
+from typing import Dict, List
+
+import camera.calib
+import images.utils as img_utils
+import process.utils as process_utils
+
+from process.skeleton_fitting import SkeletonFitting
+from images.multiimagesequence import MultiImageSequence
+from point_tracker.tracker2d import BlobDetectorSettings, BackgroundSubtractorSettings
+from point_tracker.reconstructor3d import Reconstructor3DSettings
+
+
+# import deeplabcut
+# import tensorflow as tf
+#
+# config = tf.compat.v1.ConfigProto()
+# config.gpu_options.allow_growth = True
+# config.log_device_placement = True
+# sess = tf.compat.v1.Session(config=config)
+
+
+# TODO remove specifics dependencies to insect body and wings modules => make more flexible
+from process.dlc_postprocessing import load_dlc_from_full_pickle, save_dlc2d_points_in_csv, save_dlc3d_points_in_csv, \
+ unscramble_views2d_dlc, save_stroboscopic_images, reverse_processes_2d_points, recon3d_dlc
+from process.images_preprocessing import crop_all_images, crop_image, stitch_all_images, stitch_images_multiprocessing, save_avi
+from process.point_tracking import reconstruct_3d_tracks, track_2d_objects, interp_2d_coords_of_objs
+from process.skeleton_fitting import load_skeleton2d_from_csv, load_skeleton3d_from_csv
+
+from skeleton_fitter.animals.fly import *
+from images.read import *
+
+
+class BatchProcessing:
+ """
+ Class used to process batches of recordings (each consisting of sequences of images).
+
+ Can be initialised using from_directory_by_dates or from_directory_by_names. Images can be pre-processed
+ using functions like sample_images, track2d, recon3d, crop_images and stitch_images. The recordings can be
+ saved to .avi using save_avi, and analysed using analyse_dlc (deeplabcut). Then the tracking results from
+ deeplabcut can be loaded using load_dlc and a 3d skeleton can be fitted to these results using fit_skeleton.
+
+ """
+ def __init__(self, recordings: List[MultiImageSequence], save_path: str, nb_cam: int = 2) -> None:
+ """
+
+ Args:
+ recordings: A sequence of MultiImageSequence consisting of sequences of images for each camera
+ save_path: Path to the directory where the results of the processing will be saved.
+ nb_cam: The number of camera (need to be superior or equal to 2)
+ """
+
+ self.recordings = recordings # type: List[MultiImageSequence]
+ self.nb_recordings = len(self.recordings) # type: int
+ self.save_path = save_path # type: str
+
+ if nb_cam < 2:
+ raise ValueError('The number of camera (nb_cam) should be superior or equal to 2')
+ assert all([nb_cam == r.nb_sequences for r in self.recordings])
+ self.nb_cam = nb_cam # type: int
+
+ self.recording_names = [{camn: os.path.basename(os.path.normpath(r.get_path(r.names[camn-1])))
+ for camn in range(1, self.nb_cam + 1)} for r in self.recordings] # type: List[Dict[str: str]]
+
+ self.recording_paths = [{camn: os.path.dirname(r.get_path(r.names[camn-1]))
+ for camn in range(1, self.nb_cam + 1)} for r in self.recordings] # type: List[Dict[str: str]]
+
+ self.main_camn = 1 # type: int
+
+ @staticmethod
+ def from_directory_by_dates(recording_paths: List[str], frame_rate: int, save_path: str = None,
+ type_image: str = "grey", expr: str = r"\d{8}_\d{6}",
+ format_date_str: str ='%Y%m%d_%H%M%S', threshold_s: int = 25,
+ move_to_delete: bool = False) -> "BatchProcessing":
+ """
+ Initiate BatchProcessing by gathering recording files from directory.
+ It will match the recording folders for all cameras by checking the recording dates in their name.
+
+ Args:
+ recording_paths: List of paths to directories containing each with a recording sequence.
+ frame_rate: The frame_rate of all recordings (we assume it is the same for all).
+ save_path: Path to the directory where the results of the processing will be saved.
+ If None, this path will be in the first recording path.
+ expr: Regular expression pattern matching recording folders with date.
+ format_date_str: format of the date and time in the folder names.
+ threshold_s: Will check if the date in the folder names differ from less than threshold in second
+ type_image: The type of the images ('rgb', 'grey' or 'bw').
+ move_to_delete: Whether or not to move the unmatched recording folders to a to_delete folder
+ (will be in the same parent directory as the recording folders)
+ """
+
+ if save_path is None:
+ save_path = recording_paths[0]
+
+ nb_cam = len(recording_paths)
+ cam_names = [camn for camn in range(1, nb_cam +1)]
+
+ matched_folder_names = process_utils.match_recordings_by_date(recording_paths, expr=expr,
+ format_date_str=format_date_str,
+ threshold_s=threshold_s)
+
+ if move_to_delete:
+ un_matched_folder_names = process_utils.get_unmatched_recordings(recording_paths, matched_folder_names)
+
+ if un_matched_folder_names:
+ for ind_cam in range(0, len(recording_paths)):
+ destination_path = os.path.join(recording_paths[ind_cam], 'to_delete')
+
+ if not os.path.exists(destination_path):
+ os.makedirs(destination_path)
+
+ for i in range(0, len(un_matched_folder_names)):
+ if un_matched_folder_names[i]:
+ os.rename(os.path.join(recording_paths[ind_cam], un_matched_folder_names[i][ind_cam]),
+ os.path.join(destination_path, un_matched_folder_names[i][ind_cam]))
+
+ recordings = []
+ for i in range(0, len(matched_folder_names)):
+ matched_directories = [os.path.join(recording_paths[ind_cam], f) for ind_cam, f in enumerate(matched_folder_names[i])]
+ recordings.append(MultiImageSequence.from_directory(matched_directories, frame_rate, cam_names, expr,
+ type_image=type_image))
+
+ return BatchProcessing(recordings, save_path, nb_cam)
+
+ @staticmethod
+ def from_directory_by_names(recording_paths: List[str], frame_rate: int, save_path: str = None,
+ type_image: str = "grey", threshold_ratio_diff: float = None,
+ threshold_nb_diff_char: int = None, move_to_delete: bool = True) -> "BatchProcessing":
+ """
+ Initiate BatchProcessing by gathering recording files from directory.
+ It will match the recording folders for all cameras by checking their names.
+
+ Args:
+ recording_paths: List of paths to directories containing each with a recording sequence.
+ frame_rate: The frame_rate of all recordings (we assume it is the same for all).
+ save_path: Path to the directory where the results of the processing will be saved.
+ If None, this path will be in the first recording path.
+ threshold_ratio_diff: Will check the similarity percentage (between 0 and 1)
+ threshold_nb_diff_char: Will check how many characters are different between the file names.
+ type_image: The type of the images ('rgb', 'grey' or 'bw').
+ move_to_delete: Whether or not to move the unmatched recording folders to a to_delete folder
+ (will be in the same parent directory as the recording folders)
+ """
+
+ if save_path is None:
+ save_path = recording_paths[0]
+
+ nb_cam = len(recording_paths)
+ cam_names = [camn for camn in range(1, nb_cam +1)]
+
+ matched_folder_names = process_utils.match_recordings_by_name(recording_paths, threshold_ratio_diff=threshold_ratio_diff,
+ threshold_nb_diff_char=threshold_nb_diff_char)
+ if move_to_delete:
+ un_matched_folder_names = process_utils.get_unmatched_recordings(recording_paths, matched_folder_names)
+
+ if un_matched_folder_names:
+ for ind_cam in range(0, len(recording_paths)):
+ destination_path = os.path.join(recording_paths[ind_cam], 'to_delete')
+
+ if not os.path.exists(destination_path):
+ os.makedirs(destination_path)
+
+ for i in range(0, len(un_matched_folder_names)):
+ if un_matched_folder_names[i]:
+ os.rename(os.path.join(recording_paths[ind_cam], un_matched_folder_names[i][ind_cam]),
+ os.path.join(destination_path, un_matched_folder_names[i][ind_cam]))
+
+ recordings = []
+ for i in range(0, len(matched_folder_names)):
+ matched_directories = [os.path.join(recording_paths[ind_cam], f) for ind_cam, f in enumerate(matched_folder_names[i])]
+ recordings.append(MultiImageSequence.from_directory(matched_directories, frame_rate, cam_names, type_image=type_image))
+
+ return BatchProcessing(recordings, save_path, nb_cam)
+
+ def multi_processes(self, yaml_path: str, **kwargs: Dict[str, any]) -> None:
+ """
+ Running multiple processes in parallel (much faster tha one by one) from yaml file
+
+ Args:
+ yaml_path: path of the .yaml file containing the parameters of the multiples processes to run in parallel
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['yaml_path'] = yaml_path
+ self._do_batch('multi_processes', **kwargs)
+
+ def sample_images(self, step_frame: int, **kwargs: Dict[str, any]) -> None:
+ """
+ Copy samples of the images sequences for all recordings
+
+ Args:
+ step_frame: The number of images between samples to take
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['step_frame'] = step_frame
+ self._do_batch('sample', **kwargs)
+
+ def resize_images(self, resize_ratio: float, **kwargs: Dict[str, any]) -> None:
+ """
+ Up/Down-sizing of the images from a resize_ratio
+
+ Args:
+ resize_ratio: Ratio for up- or down-sizing of the images
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['resize_ratio'] = resize_ratio
+ self._do_batch('resize', **kwargs)
+
+ def enhance_images(self, fn_name: str, factor: float, **kwargs: Dict[str, any]) -> None:
+ """
+
+ Args:
+ fn_name: 'contrast', 'brightness', 'sharpness' or 'color'
+ factor: enhancing factor: 1.0 always returns a copy of the original image,
+ lower factors mean less color (brightness, contrast, etc), and higher values more
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['factor'] = factor
+ if fn_name == 'contrast': # Enhancing contrast of images
+ self._do_batch('enhance_contrast', **kwargs)
+
+ elif fn_name == 'enhance_brightness': # Enhancing brightness of images
+ self._do_batch('enhance_color', **kwargs)
+
+ elif fn_name == 'sharpness': # Enhancing sharpness of images
+ self._do_batch('enhance_color', **kwargs)
+
+ elif fn_name == 'color': # Enhancing color of images
+ self._do_batch('enhance_color', **kwargs)
+
+ else:
+ raise ValueError("fn_name unknown. It has to be either 'contrast', 'brightness', 'sharpness' or 'color'")
+
+ def auto_contrast_images(self, **kwargs: Dict[str, any]) -> None:
+ """
+ Will run an auto-contrasting algorythm on each image
+
+ Args:
+ resize_ratio: Ratio for up- or down-sizing of the images
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ self._do_batch('autocontrast', **kwargs)
+
+ def equalize_images(self, **kwargs: Dict[str, any]) -> None:
+ """
+ Will equalize all images of each sequence together
+
+ Args:
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ self._do_batch('equalize', **kwargs)
+
+ def crop_images(self, from_fn_name: str, height: int, width: int, **kwargs: Dict[str, any]) -> None:
+ """
+ Crop images at the given height and width around previously tracked 2d coordinates in pixels
+
+ Args:
+ from_fn_name: the name of the previous process, which result-folder (/_Process/_???)
+ contains the *-*-2d_points.csv file will the tracking results that will be used for this cropping
+ height: height to which the images will each be cropped
+ width: width to which the images will each be cropped
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['from_fn_name'] = from_fn_name
+ kwargs['height_crop'], kwargs['width_crop'] = height, width
+ self._do_batch('crop', **kwargs)
+
+ def rotate_images(self, degrees: int, **kwargs: Dict[str, any]) -> None:
+ """
+ Rotating cameras views of a given number of degrees
+
+ Args:
+ degrees: Angle at which the images will be rotated counter clockwise around its centre
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['degrees'] = degrees
+ self._do_batch('rotate', **kwargs)
+
+ def stitch_images(self, **kwargs: Dict[str, any]) -> None:
+ """
+ Stitch all cameras views together to make one image per frame
+
+ Args:
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ self._do_batch('stitch', **kwargs)
+
+ def track2d(self, back_subtractor_settings: BackgroundSubtractorSettings = None,
+ blob_detector_settings: BlobDetectorSettings = None, **kwargs: Dict[str, any]) -> None:
+ """
+ Track 2d coordinates of blobs detected on the images using background subtraction
+
+ Args:
+ back_subtractor_settings: BackgroundSubtractorSettings class
+ blob_detector_settings: BlobDetectorSettings class
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['back_subtractor_settings'] = back_subtractor_settings if back_subtractor_settings is not None else BackgroundSubtractorSettings()
+ kwargs['blob_detector_settings'] = blob_detector_settings if blob_detector_settings is not None else BlobDetectorSettings()
+ self._do_batch('track2d', **kwargs)
+
+ def recon3d(self, dlt_path: str, reconstructor_settings: Reconstructor3DSettings = None, **kwargs: Dict[str, any]) -> None:
+ """
+ Reconstructing 3d tracks from 2d coordinates (previously computed using track2d)
+
+ Args:
+ dlt_path: path of the .csv file containing th 11*nb_cam DLT coefficients (3d calibration)
+ reconstructor_settings: Reconstructor3DSettings class
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['dlt_path'] = dlt_path
+ kwargs['reconstructor_settings'] = reconstructor_settings if reconstructor_settings is not None else Reconstructor3DSettings()
+ self._do_batch('recon3d', **kwargs)
+
+ def save_tiff(self, **kwargs: Dict[str, any]) -> None:
+ """
+ Save images as 8 bits .tiff
+
+ Args:
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ self._do_batch('save_tiff', **kwargs)
+
+ def save_stroboscopic_image(self, back_subtractor_settings: BackgroundSubtractorSettings = None,
+ blob_detector_settings: BlobDetectorSettings = None, **kwargs: Dict[str, any]) -> None:
+ """
+ Generate stroboscopic images (composite image base on sampled images and 2d track of the objects)
+ Then save one stroboscopic image per object, per view and per recording
+
+ Args:
+ back_subtractor_settings: BackgroundSubtractorSettings class
+ blob_detector_settings: BlobDetectorSettings class
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['back_subtractor_settings'] = back_subtractor_settings if back_subtractor_settings is not None else BackgroundSubtractorSettings()
+ kwargs['blob_detector_settings'] = blob_detector_settings if blob_detector_settings is not None else BlobDetectorSettings()
+ self._do_batch('save_strobe', **kwargs)
+
+ def save_avi(self, frame_rate: int = 24, lossless: bool = False, **kwargs: Dict[str, any]) -> None:
+ """
+ Save the recordings as .avi
+
+ Args:
+ frame_rate: the frame rate at which the video will be saved
+ lossless: Whether or not to save lossless video (e.g. not compressed)
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['frame_rate'] = frame_rate
+ kwargs['lossless'] = lossless
+ self._do_batch('save_avi', **kwargs)
+
+ def analyse_dlc(self, cfg_path: str, shuffle: int, trainingset_index: int, batch_size: int, model_name: str,
+ save_avi: bool = False, **kwargs: Dict[str, any]) -> None:
+ """
+ Analysing videos with DeepLabCut
+
+ Args:
+ cfg_path:
+ shuffle:
+ trainingset_index:
+ batch_size:
+ save_avi:
+ model_name:
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['fg_path'] = cfg_path
+ kwargs['shuffle'] = shuffle
+ kwargs['trainingset_index'] = trainingset_index
+ kwargs['batch_size'] = batch_size
+ kwargs['save_avi'] = save_avi
+ kwargs['model_name'] = model_name
+ self._do_batch('analyse_dlc', **kwargs)
+
+ def load_dlc(self, model_name: str, tracker_method: str, dlt_path: str, **kwargs: Dict[str, any]) -> None:
+ """
+ Load points from DeepLabCut (unscramble, reconstruct3d ...
+
+ Args:
+ model_name:
+ tracker_method:
+ dlt_path: path of the .csv file containing th 11*nb_cam DLT coefficients (3d calibration)
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['model_name'] = model_name
+ kwargs['tracker_method'] = tracker_method
+ kwargs['dlt_path'] = dlt_path
+ self._do_batch('load_dlc', **kwargs)
+
+ def fit_skeleton(self, animal_name: str, model_name: str, res_method: str, opt_method: str, dlt_path: str, **kwargs: Dict[str, any]) -> None:
+ """
+ Optimizing fit btw skeletons to get transformation/rotation parameters
+
+ Args:
+ animal_name:
+ model_name:
+ res_method:
+ opt_method:
+ dlt_path: path of the .csv file containing th 11*nb_cam DLT coefficients (3d calibration)
+
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ # TODO check parameter needed for fit_skeleton, shouldn't require body or wing specific lits
+ # csv_path=csv_wings_geo_path, body_param_names=body_param_names, wing_param_names=wing_param_names,
+
+ kwargs['animal_name'] = animal_name
+ kwargs['model_name'] = model_name
+ kwargs['res_method'], kwargs['opt_method'] = res_method, opt_method
+ kwargs['dlt_path'] = dlt_path
+ self._do_batch('fit_skeleton', **kwargs)
+
+ def plot_skeleton(self, **kwargs: Dict[str, any]) -> None:
+ """
+ Plot (or save) skeleton on raw images
+
+ Args:
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ self._do_batch('plot_skeleton', **kwargs)
+
+ def delete(self, to_del_names: Dict[str, List[str]], **kwargs: Dict[str, any]) -> None:
+ """
+ Move list of recording folder in a 'to_delete' folder
+
+ Args:
+ to_del_names:
+ Optional arguments:
+ rec_names_to_process: List[str]
+ camn_to_process: List[int]
+ from_fn_name: str (to continue from previous process)
+ delete_previous: bool = False
+ """
+
+ kwargs['to_del_names'] = to_del_names
+ self._do_batch('delete', **kwargs)
+
+ def _do_batch(self, fn_name: str, **kwargs: Dict[str, any]) -> None:
+ """
+ Function used to run batch processing on all the recordings
+
+ Args:
+ fn_name: Name of the function to run the batch on (e.g. 'track2d', 'crop', 'analyse_dlc', 'fit_skeleton')
+ **kwargs: A dictionary with the process parameters
+ """
+
+ start = time.time()
+
+ step_frame = 1 if 'step_frame' not in kwargs.keys() else kwargs['step_frame']
+ show_plot = False if 'show_plot' not in kwargs.keys() else kwargs['show_plot']
+ delete_previous = 'delete_previous' in kwargs.keys() and kwargs['delete_previous']
+
+ kwargs['camn_to_process'] = list(range(1, self.nb_cam +1)) if 'camn_to_process' not in kwargs.keys() else kwargs['camn_to_process']
+
+ if fn_name in ['autocontrast', 'equalize']:
+ kwargs['factor'] = None
+
+ if fn_name in 'analyse_dlc':
+ if 'trainingset_index' not in kwargs.keys(): kwargs['trainingset_index'] = 0
+ if 'batch_size' not in kwargs.keys(): kwargs['batch_size'] = 2
+
+ if fn_name in ['fit_skeleton', 'plot_skeleton']:
+ assert 1 == 0 # TODO run init of SkeletonFitting here
+
+ self._print_process(fn_name, kwargs)
+ process_path = self._get_path_process(fn_name)
+ from_fn_path = self._get_path_process(kwargs['from_fn_name']) if 'from_fn_name' in kwargs.keys() else self._get_path_process('')
+ if process_path is None: process_path = from_fn_path
+
+ main_rec_names = [names[self.main_camn] for names in self.recording_names]
+ if 'rec_names_to_process' in kwargs.keys():
+ main_rec_names = [name for name in main_rec_names if name in kwargs['rec_names_to_process']]
+ del kwargs['rec_names_to_process']
+
+ # Main process loop: Will do each process step one at a time for each recording
+ for ind_rec, main_rec_name in enumerate(main_rec_names):
+ start_rec = time.time()
+ will_delete_rec = False
+
+ recording = self.recordings[ind_rec]
+ rec_names = self.recording_names[ind_rec]
+ rec_paths = self.recording_paths[ind_rec]
+
+ frames = {camn: recording.get(camn).get_numbers() for camn in range(1, self.nb_cam + 1)}
+
+ if 'from_fn_name' in kwargs.keys():
+ image_paths = {camn: os.path.join(from_fn_path, rec_names[self.main_camn],
+ 'cam{0}'.format(camn)) for camn in range(1, self.nb_cam + 1)}
+ image_names = {camn: process_utils.get_image_names_in_directory(image_paths[camn])
+ for camn in range(1, self.nb_cam + 1)}
+
+ process_dict = yaml.load(open(os.path.join(from_fn_path, rec_names[self.main_camn],
+ rec_names[self.main_camn] + '-process_history.yaml')))
+
+ else:
+ image_names = {camn: recording.get(camn).get_names() for camn in range(1, self.nb_cam + 1)}
+ image_paths = {camn: recording.get(camn).get_path() for camn in range(1, self.nb_cam + 1)}
+
+ process_dict = self._get_init_process_dict(recording, rec_paths, frames)
+
+ if fn_name == 'multi_processes':
+ print('> Processing {0}...'.format(main_rec_name))
+ multi_processes_dict = yaml.load(open(os.path.join(kwargs['yaml_path'])))
+ multi_processes_dict[0] = process_dict[0]
+
+ self._multi_processes(ind_rec, multi_processes_dict, show_plot)
+
+ print('> {0} as been processed (time elapsed: {1:.4f} s)'.format(main_rec_name, time.time() - start_rec))
+ continue
+
+ save_paths = {camn: os.path.join(process_path, rec_names[self.main_camn],
+ 'cam{0}'.format(camn)) for camn in range(1, self.nb_cam + 1)}
+
+ cpt = 1
+ while cpt in process_dict.keys(): cpt += 1
+ process_dict[cpt] = {'fn': fn_name, 'kwargs': process_utils.convert_kwargs_to_be_writen_in_yaml(kwargs)}
+
+ if fn_name == 'stitch':
+ stitch_all_images(image_names, image_paths, frames, save_paths, self.main_camn)
+
+ elif fn_name == 'recon3d':
+ pts_csv = {}
+ for camn in range(1, self.nb_cam + 1):
+ pts_csv[camn] = np.genfromtxt(os.path.join(save_paths[camn], rec_names[camn] + '-2d_points.csv'),
+ delimiter=',', skip_header=0, names=True)
+
+ recon_sets = Reconstructor3DSettings() if 'reconstructor_settings' not in kwargs.keys() \
+ else kwargs['reconstructor_settings']
+ reconstruct_3d_tracks(pts_csv, kwargs['dlt_path'], rec_names, save_paths,
+ show_plot=show_plot, reconstructor_settings=recon_sets)
+
+ elif fn_name == 'analyse_dlc':
+ video_paths = glob.glob(os.path.join(image_paths[self.main_camn], '*.avi'))
+
+ if not len(video_paths) == 0:
+ deeplabcut.analyze_videos(kwargs['cfg_path'], video_paths, destfolder=save_paths[self.main_camn],
+ shuffle=kwargs['shuffle'], trainingset_index=kwargs['trainingset_index'],
+ batch_size=kwargs['batch_size'], save_as_csv=True)
+
+ if 'save_avi' in kwargs.keys() and kwargs['save_avi']:
+ deeplabcut.create_video_with_all_detections(save_paths[self.main_camn], video_paths,
+ kwargs['model_name'], destfolder=save_paths[self.main_camn])
+
+ elif fn_name == 'load_dlc':
+ # TODO Give options to load from csv, pickle or hdf5?
+ # skeleton2d = self.load_dlc_from_csv(kwargs['model_name'], image_paths[self.main_camn], kwargs['tracker_method'], self.nb_cam)
+ # skeleton2d = self.load_dlc_from_hdf5(kwargs['model_name'], image_paths[self.main_camn], kwargs['tracker_method'], self.nb_cam)
+ skeleton2d = load_dlc_from_full_pickle(kwargs['model_name'], image_paths[self.main_camn])
+
+ #self.test_unwrap(kwargs['dlt_path'])
+ unscramble_views2d_dlc(skeleton2d, process_dict)
+ reverse_processes_2d_points(skeleton2d, process_dict, self.threshold_likelihood, image_sizes[camn], self.save_path)
+ save_dlc2d_points_in_csv(skeleton2d, save_paths)
+
+ #self.unscramble_sides2d_and_recon3d_dlc(kwargs['dlt_path'])
+ recon3d_dlc(skeleton2d, kwargs['dlt_path'])
+ save_dlc3d_points_in_csv(skeleton3d, save_paths)
+
+ elif fn_name == 'fit_skeleton':
+ skeleton_fitting = SkeletonFitting()
+
+ csv_paths = glob.glob(os.path.join(image_paths[self.main_camn], '*-3d_points.csv'))
+ if not len(csv_paths) == 0:
+ # TODO load in SkeletonFitting? currently the skeletons are not used
+ skeleton3d = load_skeleton3d_from_csv(kwargs['model_name'], image_paths, self.main_camn)
+ skeleton2d = load_skeleton2d_from_csv(kwargs['model_name'], image_paths, self.main_camn)
+
+ if 'angles_to_correct' not in kwargs.keys(): kwargs['angles_to_correct'] = []
+ skeleton_fitting.optim_fit_skeleton(kwargs['animal_name'], kwargs['body_param_names'],
+ kwargs['wing_param_names'], kwargs['res_method'], kwargs['opt_method'],
+ rec_paths, save_paths, kwargs['dlt_path'])
+
+ elif fn_name == 'plot_skeleton':
+ skeleton_fitting = SkeletonFitting()
+
+ csv_paths = glob.glob(os.path.join(image_paths[self.main_camn], '*-2d_points.csv'))
+ if not len(csv_paths) == 0:
+ load_skeleton2d_from_csv(kwargs['model_name'], image_paths, self.main_camn)
+ body_params, wings_params = skeleton_fitting.load_params_from_csv(image_paths[self.main_camn])
+
+ dlt_coefs = camera.calib.read_dlt_coefs(kwargs['dlt_path'])
+ skeleton_fitting.plot_skeleton2d(body_params, wings_params, rec_paths, dlt_coefs, modulo=step_frame,
+ save_images=True, save_paths=save_paths)
+
+ if 'save_avi' in kwargs.keys() and kwargs['save_avi']:
+ frame_rate = 24 if 'frame_rate' not in kwargs.keys() else kwargs['frame_rate']
+ lossless = False if 'lossless' not in kwargs.keys() else kwargs['lossless']
+
+ save_avi(image_names[camn], image_paths[camn], rec_names[camn], save_paths[camn],
+ frame_rate, step_frame, lossless)
+
+ elif will_delete_rec:
+ rec_names_to_del = {camn: recording.get(camn).get_names() for camn in range(1, self.nb_cam + 1)}
+ rec_paths_to_del = {camn: recording.get(camn).get_path() for camn in range(1, self.nb_cam + 1)}
+ self._delete_recordings(rec_names_to_del, rec_paths_to_del)
+
+ else:
+ for camn in range(1, self.nb_cam + 1):
+ if will_delete_rec or camn not in kwargs['camn_to_process']: continue
+
+ if delete_previous:
+ if os.path.exists(save_paths[camn]) and rec_paths[camn] != save_paths[camn]:
+ shutil.rmtree(save_paths[camn], ignore_errors=True)
+
+ if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
+
+ image_sizes = process_dict[0]['image_sizes']
+
+ if fn_name == 'track2d':
+ back_sub_sets = BackgroundSubtractorSettings() if 'back_subtractor_settings' not in kwargs.keys() \
+ else kwargs['back_subtractor_settings']
+ blob_det_sets = BlobDetectorSettings() if 'blob_detector_settings' not in kwargs.keys() \
+ else kwargs['blob_detector_settings']
+
+ track_2d_objects(recording.get(camn), rec_names[camn], save_paths[camn], show_plot=show_plot,
+ back_subtractor_settings=back_sub_sets, blob_detector_settings=blob_det_sets)
+
+ elif fn_name == 'crop':
+ coords_objs = interp_2d_coords_of_objs(frames[camn], image_paths[camn], save_paths[camn],
+ image_sizes[camn], show_plot=show_plot)
+ assert coords_objs, "Could not find any 3d track"
+
+ crop_all_images(coords_objs, image_names[camn], image_paths[camn], save_paths[camn],
+ kwargs['height_crop'], kwargs['width_crop'])
+
+ elif fn_name == 'save_strobe':
+ radius = 50 if 'radius' not in kwargs.keys() else kwargs['radius']
+ shape = 'rectangle' if 'shape' not in kwargs.keys() else kwargs['shape']
+
+ coords_objs = interp_2d_coords_of_objs(frames[camn], image_paths[camn], save_paths[camn],
+ image_sizes[camn], show_plot=show_plot)
+ assert coords_objs, "Could not find any 3d track"
+
+ back_sub_sets = BackgroundSubtractorSettings() if 'back_subtractor_settings' not in kwargs.keys() \
+ else kwargs['back_subtractor_settings']
+ blob_det_sets = BlobDetectorSettings() if 'blob_detector_settings' not in kwargs.keys() \
+ else kwargs['blob_detector_settings']
+
+ save_stroboscopic_images(recording.get(camn), frames[camn], coords_objs, image_names[camn][0],
+ save_paths[camn], radius, shape, step_frame,
+ back_subtractor_settings=back_sub_sets,
+ blob_detector_settings=blob_det_sets)
+
+ elif fn_name == 'save_avi':
+ if os.path.exists(image_paths[camn]):
+ frame_rate = 24 if 'frame_rate' not in kwargs.keys() else kwargs['frame_rate']
+ lossless = False if 'lossless' not in kwargs.keys() else kwargs['lossless']
+
+ save_avi(image_names[camn], image_paths[camn], rec_names[camn], save_paths[camn],
+ frame_rate, step_frame, lossless)
+
+ elif fn_name == 'delete':
+ if rec_names[camn] in kwargs['to_del_names']:
+ will_delete_rec = True
+
+ else:
+ for image_name in image_names[camn]:
+ img = read_image(os.path.join(image_paths[camn], image_name))
+ image_save_name = image_name
+
+ if fn_name == 'rotate':
+ img = img.rotate(kwargs['degrees'])
+
+ elif fn_name == 'resize':
+ img = img_utils.resize(img, kwargs['resize_ratio'])
+
+ elif fn_name in ['enhance_contrast', 'enhance_brightness', 'enhance_sharpness', 'enhance_color',
+ 'autocontrast', 'equalize']:
+ img = img_utils.enhance(img, fn_name, kwargs['factor'])
+
+ elif fn_name == 'save_tiff':
+ img = Image.open(os.path.join(image_paths[camn], image_name))
+ img_utils.save_tiff(img, os.path.join(save_paths[camn], image_name))
+
+ if fn_name in ['rotate', 'resize', 'sample', 'enhance_contrast', 'enhance_brightness',
+ 'enhance_sharpness', 'enhance_color', 'autocontrast', 'equalize']:
+ img.save(os.path.join(save_paths[camn], image_save_name))
+ img.close()
+
+ if os.path.exists(save_paths[camn]) and not os.listdir(save_paths[camn]): # if save_path empty
+ shutil.rmtree(save_paths[camn], ignore_errors=True)
+
+ # Save process history in yaml file
+ self._write_process_dict(process_dict, rec_names[self.main_camn], os.path.join(process_path, rec_names[self.main_camn]))
+
+ print('> {0} as been processed (time elapsed: {1:.4f} s)'.format(main_rec_name, time.time() - start_rec))
+
+ print('> Full batch as been processed (total time elapsed: {0:.4f} s)'.format(time.time() - start))
+
+ def _multi_processes(self, ind_rec: int, multi_processes_dict: Dict[int, Dict[str, any]], show_plot=False) -> None:
+ image_sizes = multi_processes_dict[0]['image_sizes']
+
+ img, obj_names = {}, []
+ multi_processes_nums = list(multi_processes_dict.keys())
+ for multi_processes_num in multi_processes_nums[1:]:
+ start = time.time()
+ fn_name = multi_processes_dict[multi_processes_num]['fn']
+ kwargs = multi_processes_dict[multi_processes_num]['kwargs']
+ step_frame = 1 if 'step_frame' not in kwargs.keys() else kwargs['step_frame']
+ kwargs['camn_to_process'] = list(range(1, self.nb_cam + 1) if 'camn_to_process' not in kwargs.keys()
+ else kwargs['camn_to_process'])
+ recording = self.recordings[ind_rec]
+ rec_names = self.recording_names[ind_rec]
+ rec_paths = self.recording_paths[ind_rec]
+
+ from_fn_path = self._get_path_process(kwargs['from_fn_name']) if 'from_fn_name' in kwargs.keys() \
+ else self._get_path_process('')
+ process_path = self._get_path_process(fn_name)
+ if process_path is None: process_path = from_fn_path
+
+ image_names = {camn: self.recordings[ind_rec].get(camn).get_names() for camn in range(1, self.nb_cam + 1)}
+ image_paths = {camn: recording.get(camn).get_path() for camn in range(1, self.nb_cam + 1)}
+
+ save_paths = {camn: os.path.join(process_path, rec_names[self.main_camn],
+ 'cam{0}'.format(camn)) for camn in range(1, self.nb_cam + 1)}
+
+ frames = {camn: recording.get(camn).get_numbers() for camn in range(1, self.nb_cam + 1)}
+
+ for camn in range(1, self.nb_cam + 1):
+ if 'delete_previous' in kwargs.keys() and kwargs['delete_previous']:
+ if os.path.exists(save_paths[camn]) and rec_paths[camn] != save_paths[camn]:
+ shutil.rmtree(save_paths[camn], ignore_errors=True)
+
+ for camn in range(1, self.nb_cam +1):
+
+ if not obj_names or 'obj0' in obj_names:
+ file_names = sorted(os.listdir(image_paths[camn]), key=lambda s: s[:s.rfind('.')])
+ csv_names = [s for s in file_names if '-2d_points.csv' in s and '-obj' in s]
+
+ if csv_names:
+ obj_names = np.unique([s[process_utils.find_char(s, '-')[0] + 1:process_utils.find_char(s, '-')[1]] for s in csv_names])
+ if 'obj0' in img[camn].keys(): img[camn] = {}
+ else:
+ obj_names = ['obj0']
+
+ if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
+ if not camn in kwargs['camn_to_process']: continue
+ if fn_name in ['recon3d', 'stitch']: continue
+
+ if fn_name == 'track2d':
+ back_sub_sets = BackgroundSubtractorSettings() if 'back_subtractor_settings' not in kwargs.keys() \
+ else BackgroundSubtractorSettings.from_dict(kwargs['back_subtractor_settings'])
+ blob_det_sets = BlobDetectorSettings() if 'blob_detector_settings' not in kwargs.keys() \
+ else BlobDetectorSettings.from_dict(kwargs['blob_detector_settings'])
+
+ track_2d_objects(recording.get(camn), rec_names[camn], save_paths[camn],
+ back_subtractor_settings=back_sub_sets,
+ blob_detector_settings=blob_det_sets, show_plot=show_plot)
+ continue
+
+ elif fn_name == 'crop':
+ csv_path = os.path.join(from_fn_path, rec_names[self.main_camn], 'cam{0}'.format(camn))
+ coords_objs = interp_2d_coords_of_objs(frames[camn], csv_path, save_paths[camn],
+ image_sizes[camn], show_plot=show_plot)
+ assert coords_objs, "Could not find any 3d track"
+ obj_names = coords_objs.keys()
+
+ elif fn_name == 'save_strobe':
+ radius = 50 if 'radius' not in kwargs.keys() else kwargs['radius']
+ shape = 'rectangle' if 'shape' not in kwargs.keys() else kwargs['shape']
+
+ csv_path = os.path.join(from_fn_path, rec_names[self.main_camn], 'cam{0}'.format(camn))
+ coords_objs = interp_2d_coords_of_objs(frames[camn], csv_path, save_paths[camn],
+ image_sizes[camn], show_plot=show_plot)
+ assert coords_objs, "Could not find any 3d track"
+
+ back_sub_sets = BackgroundSubtractorSettings() if 'back_subtractor_settings' not in kwargs.keys() else kwargs['back_subtractor_settings']
+ blob_det_sets = BlobDetectorSettings() if 'blob_detector_settings' not in kwargs.keys() else kwargs['blob_detector_settings']
+
+ save_stroboscopic_images(recording.get(camn), frames[camn], coords_objs, image_names[camn][0],
+ save_paths[camn], radius, shape, step_frame,
+ back_subtractor_settings=back_sub_sets,
+ blob_detector_settings=blob_det_sets)
+ continue
+
+ elif fn_name == 'save_avi':
+ frame_rate = 24 if 'frame_rate' not in kwargs.keys() else kwargs['frame_rate']
+ lossless = False if 'lossless' not in kwargs.keys() else kwargs['lossless']
+
+ if camn not in img.keys(): img[camn] = {}
+
+ for obj_name in obj_names:
+ if obj_name not in img[camn].keys():
+ img[camn][obj_name] = {}
+
+ #if fn_name not in ['track2d', 'recon3d'] \
+ # and obj_name == 'obj0' and 'from_fn_name' in kwargs.keys(): break
+
+ if fn_name == 'save_avi':
+ if 'obj0' == obj_name: save_name = rec_names[camn] + '.avi'
+ else: save_name = rec_names[camn] + '-' + obj_name + '.avi'
+
+ if len(np.shape(np.asarray(img[camn][obj_name][0]))) != 2: break
+ height_img, width_img = np.shape(np.asarray(img[camn][obj_name][0]))
+
+ codec = cv2.VideoWriter_fourcc(*'HFYU') if lossless else cv2.VideoWriter_fourcc(*'MP42')
+ writer = cv2.VideoWriter(os.path.join(save_paths[camn], save_name), codec, frame_rate, (width_img, height_img))
+ #continue
+
+ for ind_img, image_name in enumerate(image_names[camn]):
+ if ind_img not in img[camn][obj_name].keys():
+ img[camn][obj_name][ind_img] = read_image(os.path.join(image_paths[camn], image_name))
+
+ if not img[camn][obj_name][ind_img]: continue
+
+ if fn_name == 'crop':
+ coords_obj = {key: coords_objs[obj_name][key][ind_img] for key in coords_objs[obj_name].keys()}
+
+ img[camn][obj_name][ind_img] = crop_image(img[camn][obj_name][ind_img], coords_obj,
+ kwargs['height_crop'], kwargs['width_crop'])
+
+ elif fn_name == 'rotate':
+ img[camn][obj_name][ind_img] = img[camn][obj_name][ind_img].rotate(kwargs['degrees'])
+
+ elif fn_name == 'resize':
+ img[camn][obj_name][ind_img] = img_utils.resize(img[camn][obj_name][ind_img], kwargs['resize_ratio'])
+
+ elif fn_name in ['enhance_contrast', 'enhance_brightness', 'enhance_sharpness', 'enhance_color',
+ 'autocontrast', 'equalize']:
+ img[camn][obj_name][ind_img] = img_utils.enhance(img[camn][obj_name][ind_img], fn_name, kwargs['factor'])
+
+ elif fn_name == 'save_avi':
+ writer.write(cv2.cvtColor(np.asarray(img[camn][obj_name][ind_img]), cv2.COLOR_GRAY2BGR))
+
+ if fn_name == 'save_avi': writer.release()
+
+ if 'step_frame' in kwargs.keys() or multi_processes_num == multi_processes_nums[-1]: # Will save images
+ if fn_name in ['crop', 'rotate', 'resize', 'sample', 'enhance_contrast', 'enhance_brightness',
+ 'enhance_sharpness', 'enhance_color', 'autocontrast', 'equalize']:
+ for ind_img, image_name in enumerate(image_names[camn]):
+ if 'obj0' == obj_name: image_save_name = image_name
+ else: image_save_name = image_name[:image_name.rfind('.')] + '-' + obj_name + image_name[image_name.rfind('.'):]
+
+ img[camn][obj_name][ind_img].save(os.path.join(save_paths[camn], image_save_name))
+ #img[camn][obj_name][i] = []
+
+ if os.path.exists(save_paths[camn]) and not os.listdir(save_paths[camn]): # if save_path empty
+ shutil.rmtree(save_paths[camn], ignore_errors=True)
+
+ if fn_name == 'recon3d':
+ pts_csv = {}
+ for camn in range(1, self.nb_cam + 1):
+ pts_csv[camn] = np.genfromtxt(os.path.join(save_paths[camn], rec_names[camn] + '-2d_points.csv'),
+ delimiter=',', skip_header=0, names=True)
+
+ recon_sets = Reconstructor3DSettings() if 'reconstructor_settings' not in kwargs.keys() \
+ else Reconstructor3DSettings.from_dict(kwargs['reconstructor_settings'])
+ reconstruct_3d_tracks(pts_csv, kwargs['dlt_path'], rec_names, save_paths, show_plot=show_plot,
+ reconstructor_settings=recon_sets)
+
+ elif fn_name == 'stitch':
+ img = stitch_images_multiprocessing(img, image_names, image_paths, frames, save_paths, obj_names,
+ multi_processes_num, multi_processes_nums, self.main_camn)
+
+ # Save process history in yaml file
+ self._write_process_dict(multi_processes_dict, rec_names[self.main_camn],
+ os.path.join(process_path, rec_names[self.main_camn]))
+
+ print('>> fn {0} as been performed (time elapsed: {1:.4f} s)'.format(fn_name, time.time() - start))
+
+ @staticmethod
+ def _print_process(fn_name: str, kwargs: Dict[str, any]) -> None:
+ print('____________________________')
+ if fn_name == 'multi_processes':
+ print('> Starting batch images: Running multiple processes')
+ elif fn_name == 'sample':
+ print('> Starting batch images: Copying samples')
+ elif fn_name == 'resize':
+ print('> Starting batch images: Up/Down-sizing images (ratio: {0})'.format(kwargs['resize_ratio']))
+ elif fn_name == 'enhance_contrast':
+ print('> Starting batch images: Enhancing contrast of images (factor: {0})'.format(kwargs['factor']))
+ elif fn_name == 'enhance_brightness':
+ print('> Starting batch images: Enhancing brightness of images (factor: {0})'.format(kwargs['factor']))
+ elif fn_name == 'enhance_sharpness':
+ print('> Starting batch images: Enhancing sharpness of images (factor: {0})'.format(kwargs['factor']))
+ elif fn_name == 'enhance_color':
+ print('> Starting batch images: Enhancing color of images (factor: {0})'.format(kwargs['factor']))
+ elif fn_name == 'autocontrast':
+ print('> Starting batch images: Auto-Enhancing contrast of images')
+ elif fn_name == 'equalize':
+ print('> Starting batch images: Equalizing of images')
+ elif fn_name == 'crop':
+ print('> Starting batch images: Cropping images (height: {0}, width: {1})'.format(kwargs['height_crop'], kwargs['width_crop']))
+ elif fn_name == 'rotate':
+ print('> Starting batch images: Rotating cameras views {0} at {1} degrees '.format(kwargs['camn_to_process'], kwargs['degrees']))
+ elif fn_name == 'stitch':
+ print('> Starting batch images: Stitching all cameras views together')
+ elif fn_name == 'track2d':
+ print('> Starting batch images: Tracking 2d coordinates of blobs')
+ elif fn_name == 'recon3d':
+ print('> Starting batch images: Reconstructing 3d tracks from 2d coordinates')
+ elif fn_name == 'save_tiff':
+ print('> Starting batch images: Saving recordings as 8 bits .tiff')
+ elif fn_name == 'save_strobe':
+ print('> Starting batch images: Generating stroboscopic images from 2d coordinates')
+ elif fn_name == 'save_avi':
+ print('> Starting batch images: Saving recordings as .avi')
+ elif fn_name == 'analyse_dlc':
+ print('> Starting batch images: Analysing videos with DeepLabCut (model: {0})'.format(kwargs['model_name']))
+ elif fn_name == 'load_dlc':
+ print('> Starting batch images: Load points from DeepLabCut (unscramble, reconstruct3d ... - model: {0})'.format(kwargs['model_name']))
+ elif fn_name == 'fit_skeleton':
+ print('> Starting batch images: Optimizing fit btw skeletons to get transformation/rotation parameters')
+ elif fn_name == 'plot_skeleton':
+ print('> Starting batch images: Plot (or save) skeleton on raw images')
+ elif fn_name == 'delete':
+ print('> Starting batch images: Deleting following recordings: {0}'.format(kwargs['to_del_names']))
+
+ def _get_path_process(self, fn_name: str = '') -> str:
+
+ if fn_name == 'crop':
+ process_path = os.path.join(self.save_path, '_Cropped')
+ elif fn_name == 'sample':
+ process_path = os.path.join(self.save_path, '_Sample')
+ elif fn_name == 'resize':
+ process_path = os.path.join(self.save_path, '_Resized')
+ elif fn_name in ['enhance_contrast', 'enhance_brightness', 'enhance_sharpness', 'enhance_color', 'autocontrast', 'equalize']:
+ process_path = os.path.join(self.save_path, '_Enhanced')
+ elif fn_name == 'save_tiff':
+ process_path = os.path.join(self.save_path, '_Tiff')
+ elif fn_name == 'save_strobe':
+ process_path = os.path.join(self.save_path, '_Strobe')
+ elif fn_name == 'save_avi':
+ process_path = os.path.join(self.save_path, '_Avi')
+ elif fn_name == 'stitch':
+ process_path = os.path.join(self.save_path, '_Stitched')
+ elif fn_name in ['analyse_dlc', 'load_dlc']:
+ process_path = os.path.join(self.save_path, '_DLC')
+ elif fn_name == 'plot_skeleton':
+ process_path = os.path.join(self.save_path, '_Plots')
+
+ elif fn_name in ['', 'rotate', 'delete', 'track2d', 'recon3d', 'multi_processes', 'fit_skeleton']:
+ process_path = None
+
+ else:
+ raise ValueError('WARN: incorrect function name (fn_name = {0})'.format(fn_name))
+
+ return process_path
+
+ def _get_init_process_dict(self, recording: MultiImageSequence, rec_paths: Dict[str, str],
+ frames: Dict[str, List[int]]) -> Dict[any, any]:
+
+ process_dict, cpt = {}, 1
+
+ last_frame = int(np.max([frames[camn][-1] for camn in range(1, self.nb_cam + 1)]))
+
+ missing_frames, image_sizes = {1: {}}, {1: {}}
+ for camn in range(1, self.nb_cam + 1):
+ missing_frames[camn] = list(set(range(1, last_frame +1)) - set(frames[camn]))
+ image_sizes[camn] = Image.open(os.path.join(recording.get(camn).get_path(), recording.get(camn).get_names()[0])).size
+
+ process_dict[0] = {'fn': 'initialization', 'nb_cam': self.nb_cam,
+ 'paths': rec_paths, 'process_path': self.save_path,
+ 'image_sizes': image_sizes, 'last_frame': last_frame, 'missing_frames': missing_frames}
+
+ return process_dict
+
+ @staticmethod
+ def _write_process_dict(process_dict: Dict[int, Dict[str, any]], save_name: str, save_path: str) -> None:
+ with open(os.path.join(save_path, save_name + '-process_history.yaml'), 'w') as yaml_file:
+ yaml.safe_dump(process_dict, yaml_file, encoding='utf-8', allow_unicode=True)
+
+ @staticmethod
+ def _delete_recordings(rec_names: Dict[str, List[str]], rec_paths: Dict[str, List[str]]) -> None:
+
+ for camn in rec_paths.keys():
+
+ if not len(rec_names[camn]) == 0:
+ if not os.path.exists(os.path.join(rec_paths[camn], 'to_delete')):
+ os.makedirs(os.path.join(rec_paths[camn], 'to_delete'))
+
+ for rec_name in rec_names:
+ os.rename(os.path.join(rec_paths[camn], rec_name),
+ os.path.join(rec_paths[camn], 'to_delete', rec_name))
+ # shutil.rmtree(os.path.join(self.cam_paths[camn], rec_name), ignore_errors=True)
+
+ # print('>> recording from {0} has been deleted'.format(time.strftime('%Y%m%d_%H%M%S', rec_dates[camn])))
+
+
+if __name__ == '__main__':
+
+ # TODO update loading BatchProcessing from path
+ print('TODO')
+
+
+
+
diff --git a/process/dlc_postprocessing.py b/process/dlc_postprocessing.py
new file mode 100644
index 0000000000000000000000000000000000000000..27844d71dc31760b91e5e4492df3d51601c110f9
--- /dev/null
+++ b/process/dlc_postprocessing.py
@@ -0,0 +1,857 @@
+import copy, glob, os, pickle
+
+import pandas as pd
+import numpy as np
+import yaml
+
+import matplotlib.pyplot as plt
+from PIL import Image
+from scipy import signal
+
+from point_tracker import Reconstructor3D
+from point_tracker.reconstructor3d import Reconstructor3DSettings
+
+from point_tracker.tracker2d import Tracker2D, BlobDetectorSettings, BackgroundSubtractorSettings
+
+import skeleton_fitter.skeleton_modules.insect_body_slim as body
+
+
+def load_dlc_from_csv(model_name, tracker_method, path):
+ process_dict = yaml.load(open(glob.glob(os.path.join(path, '*-process_history.yaml'))[0]))
+ frames = list(range(1, process_dict[0]['last_frame']))
+
+ nb_cam = len(process_dict[0]['missing_frames'].keys())
+ missing_frames = list(np.unique(np.concatenate([process_dict[0]['missing_frames'][camn] for camn in range(1, nb_cam + 1)])))
+ [missing_frames.remove(missing_frame) for missing_frame in list(set(missing_frames) - set(frames))]
+ [frames.remove(missing_frame) for missing_frame in missing_frames]
+
+ if tracker_method == 'skeleton': tracker_method_str = '_sk'
+ elif tracker_method == 'box': tracker_method_str = '_bk'
+
+ dlc_csv_paths = glob.glob(os.path.join(path, '*' + model_name + tracker_method_str + '.csv'))
+
+ obj_names = []
+ for dlc_csv_path in dlc_csv_paths: # go throught the various obj
+ obj_names.append(dlc_csv_path[len(path) + 1:dlc_csv_path.index(model_name)])
+
+ skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
+ for i, obj_name in enumerate(obj_names):
+ df = pd.read_csv(dlc_csv_paths[i], header=[1, 2])
+ label_names = np.unique([s[0][:-2] for s in df.keys()[1:]])
+ new_label_names = [label.replace(' ', '_') for label in label_names]
+
+ skeleton2d[obj_name] = {'frames': frames, 'label_names': new_label_names}
+ for j, label in enumerate(label_names):
+ new_label = new_label_names[j]
+
+ skeleton2d[obj_name][new_label] = {1: {}}
+ for camn in range(1, nb_cam + 1):
+ label_key = label + ' {0}'.format(camn)
+
+ skeleton2d[obj_name][new_label][camn] = {'x': np.array(df[(label_key, 'x')]),
+ 'y': np.array(df[(label_key, 'y')]),
+ 'likelihood': np.array(df[(label_key, 'likelihood')])}
+
+ return skeleton2d
+
+
+def load_dlc_from_full_pickle(model_name, path):
+ process_dict = yaml.load(open(glob.glob(os.path.join(path, '*-process_history.yaml'))[0]))
+ frames = list(range(1, process_dict[0]['last_frame']))
+
+ nb_cam = len(process_dict[0]['missing_frames'].keys())
+ missing_frames = list(np.unique(np.concatenate([process_dict[0]['missing_frames'][camn] for camn in range(1, nb_cam + 1)])))
+ [missing_frames.remove(missing_frame) for missing_frame in list(set(missing_frames) - set(frames))]
+ [frames.remove(missing_frame) for missing_frame in missing_frames]
+
+ dlc_pickle_paths = glob.glob(os.path.join(path, '*' + model_name + '_full.pickle'))
+
+ obj_names = []
+ for dlc_pickle_path in dlc_pickle_paths: # go throught the various obj
+ obj_names.append(dlc_pickle_path[len(path) + 22:dlc_pickle_path.index(model_name)])
+
+ skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
+ for obj_name in obj_names:
+ with open(dlc_pickle_paths[0], "rb") as file:
+ full_dict = pickle.load(file)
+
+ header = full_dict.pop("metadata")
+ label_names = np.unique([s[:-2] for s in header["all_joints_names"]])
+ new_label_names = [label.replace(' ', '_') for label in label_names]
+
+ skeleton2d[obj_name] = {'frames': frames, 'label_names': new_label_names}
+ for i, label in enumerate(label_names):
+ new_label = new_label_names[i]
+
+ skeleton2d[obj_name][new_label] = {1: {}}
+ for camn in range(1, nb_cam + 1):
+ skeleton2d[obj_name][new_label][camn] = {'x': np.nan * np.ones(len(frames)),
+ 'y': np.nan * np.ones(len(frames)),
+ 'likelihood': np.nan * np.ones(len(frames))}
+
+ for frame_name in full_dict.keys():
+ try:
+ frame_in = frames.index(int(frame_name[-4:]) + 1)
+ except ValueError:
+ continue
+
+ for i, label in enumerate(label_names):
+ new_label = new_label_names[i]
+
+ for camn in range(1, nb_cam + 1):
+ try: label_in = header["all_joints_names"].index(label + ' {0}'.format(camn))
+ except ValueError: continue
+
+ if len(full_dict[frame_name]['confidence'][label_in]) == 0: continue
+ best_in = np.argmax(full_dict[frame_name]['confidence'][label_in])
+
+ skeleton2d[obj_name][new_label][camn]['x'][frame_in] = \
+ full_dict[frame_name]['coordinates'][0][label_in][best_in][0]
+ skeleton2d[obj_name][new_label][camn]['y'][frame_in] = \
+ full_dict[frame_name]['coordinates'][0][label_in][best_in][1]
+ skeleton2d[obj_name][new_label][camn]['likelihood'][frame_in] = \
+ full_dict[frame_name]['confidence'][label_in][best_in]
+
+ return skeleton2d
+
+
+def load_dlc_from_hdf5(model_name, tracker_method, path):
+ process_dict = yaml.load(open(glob.glob(os.path.join(path, '*-process_history.yaml'))[0]))
+ frames = list(range(1, process_dict[0]['last_frame']))
+
+ nb_cam = len(process_dict[0]['missing_frames'].keys())
+ missing_frames = np.unique(
+ [np.squeeze(process_dict[0]['missing_frames'][camn]) for camn in range(1, nb_cam + 1)])
+ [frames.remove(missing_frame) for missing_frame in missing_frames]
+
+ if tracker_method == 'skeleton': tracker_method_str = '_sk'
+ elif tracker_method == 'box': tracker_method_str = '_bk'
+
+ dlc_hdf5_paths = glob.glob(os.path.join(path, '*' + model_name + tracker_method_str + '.h5'))
+
+ obj_names = []
+ for dlc_hdf5_path in dlc_hdf5_paths: # go throught the various obj
+ obj_names.append(dlc_hdf5_path[len(path) + 22:dlc_hdf5_path.index(model_name)])
+
+ skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
+ for i, obj_name in enumerate(obj_names):
+ hdf5_dict = pd.read_hdf(dlc_hdf5_paths[i], 'df_with_missing')
+ # hdf5_dict.to_csv(os.path.join(save_path, hdf5_name.split(".h5")[0] + '.csv'))
+
+ # network_name = hdf5_dict.columns[0][0]
+ hdf5_dict.columns = hdf5_dict.columns.droplevel([0, 1]) # remove first two lines of header (network_name, individual)
+
+ label_names = np.unique([s[:-2] for s in hdf5_dict.columns.droplevel(1)])
+ new_label_names = [label.replace(' ', '_') for label in label_names]
+ # frames = hdf5_dict.index
+
+ skeleton2d[obj_name] = {'frames': frames, 'label_names': new_label_names}
+ for i, label in enumerate(label_names):
+ new_label = new_label_names[i]
+
+ skeleton2d[obj_name][new_label] = {1: {}}
+ for camn in range(1, nb_cam + 1):
+ label_key = label + ' {0}'.format(camn)
+
+ skeleton2d[obj_name][new_label][camn] = {'x': np.array(hdf5_dict[(label_key, 'x')]),
+ 'y': np.array(hdf5_dict[(label_key, 'y')]),
+ 'likelihood': np.array(hdf5_dict[(label_key, 'likelihood')])}
+
+ # skeleton2d[obj_name][new_label][camn] = [[1]*3 for i in range(len(hdf5_dict[(label_key, 'x')]))]
+ # # skeleton2d[obj_name][new_label][camn] = np.ones((len(hdf5_dict[(label_key, 'x')]), 3))
+ # skeleton2d[obj_name][new_label][camn][:, 0] = np.array(hdf5_dict[(label_key, 'x')])
+ # skeleton2d[obj_name][new_label][camn][:, 1] = np.array(hdf5_dict[(label_key, 'y')])
+ # skeleton2d[obj_name][new_label][camn][:, 2] = np.array(hdf5_dict[(label_key, 'likelihood')])
+
+ return skeleton2d
+
+
+def save_dlc2d_points_in_csv(skeleton2d, save_paths):
+
+ for obj_name in skeleton2d['obj_names']:
+ for label in skeleton2d[obj_name]['label_names']:
+
+ nb_cam = len(skeleton2d[obj_name][label].keys())
+ for camn in range(1, nb_cam + 1):
+ if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
+ np.savetxt(os.path.join(save_paths[camn], obj_name + '-' + label + '-2d_points.csv'),
+ np.c_[skeleton2d[obj_name]['frames'], skeleton2d[obj_name][label][camn]['x'],
+ skeleton2d[obj_name][label][camn]['y'],
+ skeleton2d[obj_name][label][camn]['likelihood']],
+ delimiter=',', header='frame,x_px,y_px,likelihood')
+
+ # # To save all in same .csv file
+ # for camn in range(1, nb_cam + 1):
+ # if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
+ # header = header + ', x{0}_px, y{0}_px, likelihood{0}'.format(camn)
+ # array_csv = np.concatenate((array_csv, np.c_[self.skeleton2d[obj_name][label][camn]['x'],
+ # self.skeleton2d[obj_name][label][camn]['y'],
+ # self.skeleton2d[obj_name][label][camn]['likelihood']), axis=1)
+ #
+ # np.savetxt(os.path.join(save_paths[self.main_camn], obj_name + '-' + label + '-2d_points.csv'),
+ # array_csv, delimiter=',', header=header)
+
+
+def save_dlc3d_points_in_csv(skeleton3d, save_paths):
+ for obj_name in skeleton3d['obj_names']:
+ for label in skeleton3d[obj_name]['label_names']:
+
+ nb_cam = len(skeleton3d[obj_name][label].keys())
+ for camn in range(1, nb_cam + 1):
+ if not os.path.exists(save_paths[camn]): os.makedirs(save_paths[camn])
+ np.savetxt(os.path.join(save_paths[camn], obj_name + '-' + label + '-3d_points.csv'),
+ np.c_[skeleton3d[obj_name]['frames'], skeleton3d[obj_name][label]['x'],
+ skeleton3d[obj_name][label]['y'], skeleton3d[obj_name][label]['z']],
+ delimiter=',', header='frame,x,y,z')
+
+
+def unscramble_views2d_dlc(skeleton2d, process_dict):
+ # Run on unconverted 2d coords from dlc
+ # shared_axis: 0 if the combination of cam share the x axis (top-left to top-right)
+ # or 1 if they share the y axis (top-left to bottom left)
+
+ process_names = [process_dict[1][process_num]['fn'] for process_num in process_dict[1].keys()]
+ crop_num = process_names.index('crop')
+ crop_num = crop_num[-1] if isinstance(crop_num, list) else crop_num
+ width_crop = process_dict[1][crop_num]['kwargs']['width_crop']
+ height_crop = process_dict[1][crop_num]['kwargs']['height_crop']
+
+ for obj_name in skeleton2d['obj_names']:
+
+ # Detect body parts in wrong view
+ for label in skeleton2d[obj_name]['label_names']:
+ is_nan, new_camns = {}, {}
+ sum_in_out = np.zeros(np.size(skeleton2d[obj_name]['frames']))
+
+ nb_cam = len(skeleton2d[obj_name][label].keys())
+ for camn in range(1, nb_cam + 1):
+ is_nan[camn] = np.isnan(skeleton2d[obj_name][label][camn]['x'])
+
+ # in_out will be 0 if the camera view is correct, otherwise an int (-1, 1, 2, etc)
+ in_out_x = np.ceil((skeleton2d[obj_name][label][camn]['x'] - (camn - 1) * width_crop) / width_crop) - 1
+ in_out_y = np.ceil((skeleton2d[obj_name][label][camn]['y']) / height_crop) - 1
+
+ new_camns[camn] = np.ones(np.size(in_out_x)) *camn
+ for i, frame in enumerate(skeleton2d[obj_name]['frames']): # assumes that stitching was done along x axis
+ if (in_out_x[i] + camn) < 1 or (in_out_x[i] + camn) > nb_cam: new_camns[camn][i] = np.nan
+ elif in_out_y[i] < 0 or in_out_y[i] > 0: new_camns[camn][i] = np.nan
+ else: new_camns[camn][i] = in_out_x[i] + camn
+
+ sum_in_out += new_camns[camn] - camn
+
+ skeleton2d_cur_bodypart = copy.deepcopy(skeleton2d[obj_name][label])
+ for camn in range(1, nb_cam + 1):
+ skeleton2d_cur_bodypart[camn]['x'] *= np.nan
+ skeleton2d_cur_bodypart[camn]['y'] *= np.nan
+
+ for i, frame in enumerate(skeleton2d[obj_name]['frames']):
+ for camn in range(1, nb_cam + 1):
+ new_camn = new_camns[camn][i]
+ if np.isnan(new_camn) or is_nan[camn][i]: continue
+
+ if sum_in_out[i] == 0 or (is_nan[new_camn][i] and sum_in_out[i] != 0):
+ skeleton2d_cur_bodypart[new_camn]['x'][i] = skeleton2d[obj_name][label][camn]['x'][i]
+ skeleton2d_cur_bodypart[new_camn]['y'][i] = skeleton2d[obj_name][label][camn]['y'][i]
+
+ elif not np.isnan(new_camns[new_camn][i]):
+ skeleton2d_cur_bodypart[new_camn]['x'][i] = skeleton2d[obj_name][label][new_camn]['x'][i]
+ skeleton2d_cur_bodypart[new_camn]['y'][i] = skeleton2d[obj_name][label][new_camn]['y'][i]
+
+ # if self.show_plot:
+ # fig, axs = plt.subplots(1, nb_cam)
+ # for i, camn in enumerate(range(1, nb_cam + 1)):
+ # axs[i].clear()
+ # axs[i].plot(self.skeleton2d[obj_name][label][camn]['x'], self.skeleton2d[obj_name][label][camn]['y'])
+ # axs[i].plot(skeleton2d_cur_bodypart[camn]['x'], skeleton2d_cur_bodypart[camn]['y'])
+ # axs[i].set_ylabel('y pos (m) - ' + label)
+ # axs[i].set_xlabel('x pos (m)')
+ # plt.show()
+
+ skeleton2d[obj_name][label] = skeleton2d_cur_bodypart
+
+
+def unscramble_sides3d_dlc(skeleton3d, skeleton2d, body_skeleton3d_init, body_params, show_plot=False):
+ # Unscramble case where side is wrong on all cam views
+ for obj_name in skeleton3d['obj_names']:
+ label_names_right = [s for s in skeleton3d[obj_name]['label_names'] if 'right' in s]
+
+ if (skeleton2d[obj_name]['frames'] != skeleton3d[obj_name]['frames']).any():
+ raise ValueError('frames should be the same for skeleton2d and skeleton3d')
+
+ if show_plot:
+ nb_cam = len(skeleton2d[obj_name][label_names_right[0]].keys())
+
+ fig1, axs1 = plt.subplots(nb_cam, 1)
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ axs1[i].clear()
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+ x_right = skeleton2d[obj_name][label_right][camn]['x']
+ y_right = skeleton2d[obj_name][label_right][camn]['y']
+ x_left = skeleton2d[obj_name][label_left][camn]['x']
+ y_left = skeleton2d[obj_name][label_left][camn]['y']
+
+ axs1[i].plot(np.arctan2(y_right - y_left, x_right - x_left)) # angle of the line btw the two sides
+ axs1[i].set_ylabel('angle btw right and left (rad)')
+ axs1[i].set_xlabel('frames - ' + label_right)
+
+ for i, frame in enumerate(skeleton3d[obj_name]['frames']):
+ body_params_frame = {}
+ for label in body_params[obj_name].keys():
+ body_params_frame[label] = body_params[obj_name][label][i].copy()
+
+ # TODO get rid of dependency to skeleton_fitter.skeleton_modules.insect_body_slim?
+ body_skeleton3d = copy.deepcopy(body_skeleton3d_init)
+ body_skeleton3d = body.build_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton3d = body.rotate_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton3d = body.translate_skeleton3d(body_skeleton3d, body_params_frame)
+ # body_skeleton2d = body.reproject_skeleton3d_to2d(body_skeleton3d, self.dlt_coefs)
+
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+
+ x_right = skeleton3d[obj_name][label_right]['x'][i].copy()
+ y_right = skeleton3d[obj_name][label_right]['y'][i].copy()
+ z_right = skeleton3d[obj_name][label_right]['z'][i].copy()
+
+ x_left = skeleton3d[obj_name][label_left]['x'][i].copy()
+ y_left = skeleton3d[obj_name][label_left]['y'][i].copy()
+ z_left = skeleton3d[obj_name][label_left]['z'][i].copy()
+
+ xyz_right = [x_right, y_right, z_right]
+ xyz_left = [x_left, y_left, z_left]
+
+ dist_right2right_hinge = np.linalg.norm(body_skeleton3d['right_wing_hinge'] - xyz_right)
+ dist_left2right_hinge = np.linalg.norm(body_skeleton3d['right_wing_hinge'] - xyz_left)
+ dist_right2left_hinge = np.linalg.norm(body_skeleton3d['left_wing_hinge'] - xyz_right)
+ dist_left2left_hinge = np.linalg.norm(body_skeleton3d['left_wing_hinge'] - xyz_left)
+
+ if dist_right2right_hinge > dist_right2left_hinge and dist_left2left_hinge > dist_left2right_hinge:
+ x_right, y_right, z_right, x_left, y_left, z_left = x_left, y_left, z_left, x_right, y_right, z_right
+
+ skeleton3d[obj_name][label_right]['x'][i], skeleton3d[obj_name][label_right]['y'][i], \
+ skeleton3d[obj_name][label_right]['z'][i] = x_right, y_right, z_right
+
+ skeleton3d[obj_name][label_left]['x'][i], skeleton3d[obj_name][label_left]['y'][i], \
+ skeleton3d[obj_name][label_left]['z'][i] = x_left, y_left, z_left
+
+ for camn in range(1, nb_cam + 1):
+ skeleton2d[obj_name][label_right][camn]['x'][i], \
+ skeleton2d[obj_name][label_right][camn]['y'][i], \
+ skeleton2d[obj_name][label_left][camn]['x'][i], \
+ skeleton2d[obj_name][label_left][camn]['y'][i] = \
+ skeleton2d[obj_name][label_left][camn]['x'][i], \
+ skeleton2d[obj_name][label_left][camn]['y'][i], \
+ skeleton2d[obj_name][label_right][camn]['x'][i], \
+ skeleton2d[obj_name][label_right][camn]['y'][i]
+
+ if show_plot:
+ fig2, axs2 = plt.subplots(nb_cam, 1)
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ axs2[i].clear()
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+ x_right = skeleton2d[obj_name][label_right][camn]['x']
+ y_right = skeleton2d[obj_name][label_right][camn]['y']
+ x_left = skeleton2d[obj_name][label_left][camn]['x']
+ y_left = skeleton2d[obj_name][label_left][camn]['y']
+
+ axs2[i].plot(
+ np.arctan2(y_right - y_left, x_right - x_left)) # angle of the line btw the two sides
+ axs2[i].set_ylabel('angle btw right and left (rad)')
+ axs2[i].set_xlabel('frames - ' + label_right)
+
+ plt.show()
+
+
+def save_stroboscopic_images(image_sequence, frames, coords_objs, save_name, save_path, radius, shape, step_frame,
+ back_subtractor_settings: BackgroundSubtractorSettings,
+ blob_detector_settings: BlobDetectorSettings,):
+
+ if len(coords_objs.keys()) == 0:
+ print('WARM: coords_objs is empty!')
+ return
+
+ tracker = Tracker2D()
+ tracker.get_back_subtractor(back_subtractor_settings)
+ tracker.get_blob_detector(blob_detector_settings)
+
+ tracker.dist2Threshold = 7
+ tracker.load_images(image_sequence.get_names(), image_sequence.get_path(), frames)
+
+ for obj_name in coords_objs.keys():
+ tracker.points = {'x': np.array(coords_objs[obj_name]['x_px']), 'y': np.array(coords_objs[obj_name]['y_px']),
+ 'frame': np.array(coords_objs[obj_name]['frame']),
+ 'area': np.ones(np.size(coords_objs[obj_name]['x_px'])) * np.nan}
+
+ tracker.gen_stroboscopic_image(radius, shape, step_frame, flip_time=True)
+ tracker.save_stroboscopic_image(save_name + '-' + obj_name, save_path)
+
+
+def unscramble_sides2d_and_recon3d_dlc(cutoff_frequency, frame_rate, skeleton2d, dlt_path,
+ reconstructor_settings: Reconstructor3DSettings, show_plot=False):
+ # TODO! check why do not work as planed or delete
+
+ reconstructor = Reconstructor3D.read_dlt_coefs(dlt_path)
+ reconstructor.settings = reconstructor_settings
+ nb_cam = reconstructor.nb_cam
+
+ # Unscamble case where side is wrong on only a minority of cam views
+ skeleton3d = {'obj_names': skeleton2d['obj_names'], 'model_name': skeleton2d['model_name']}
+ for obj_name in skeleton2d['obj_names']:
+ label_names_right = [s for s in skeleton2d[obj_name]['label_names'] if 'right' in s]
+ other_label_names = [s for s in skeleton2d[obj_name]['label_names'] if
+ 'right' not in s or 'left' not in s]
+
+ skeleton3d[obj_name] = {'frames': skeleton2d[obj_name]['frames'],
+ 'label_names': skeleton2d[obj_name]['label_names']}
+ for label in skeleton3d[obj_name]['label_names']:
+ skeleton3d[obj_name][label] = {'x': np.nan * np.ones(np.size(skeleton2d[obj_name]['frames'])),
+ 'y': np.nan * np.ones(np.size(skeleton2d[obj_name]['frames'])),
+ 'z': np.nan * np.ones(np.size(skeleton2d[obj_name]['frames']))}
+
+ # Compute angles btw the right and left wing/bodyparts
+ angles, angles_out, mean_angles = {}, {}, {}
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ angles[camn], angles_out[camn] = {}, {}
+ mean_angles[camn] = np.zeros(np.size(skeleton2d[obj_name][label_names_right[0]][camn]['x']))
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+ label = label_right.replace('right', '')
+
+ x_right = skeleton2d[obj_name][label_right][camn]['x']
+ y_right = skeleton2d[obj_name][label_right][camn]['y']
+ x_left = skeleton2d[obj_name][label_left][camn]['x']
+ y_left = skeleton2d[obj_name][label_left][camn]['y']
+
+ angles[camn][label] = np.arctan2(y_right - y_left, x_right - x_left)
+ #angles[camn][label] = np.arctan((y_right - y_left) / (x_right - x_left))
+
+ angles[camn][label] = abs(angles[camn][label]) # doesn't not work when pi jump btw pi/2 and -pi/2
+
+ # angles[camn][label] = np.fmod(angles[camn][label], 2 * np.pi)
+ # angles[camn][label][~np.isnan(angles[camn][label])] = \
+ # np.unwrap(angles[camn][label][~np.isnan(angles[camn][label])], discont=2 * np.pi)
+ # # angles[camn][label][~np.isnan(angles[camn][label])] = \
+ # # np.unwrap(angles[camn][label][~np.isnan(angles[camn][label])] * 2, discont=2 * np.pi * 0.8) / 2
+
+ is_nan, x = np.isnan(angles[camn][label]), lambda z: z.nonzero()[0]
+ if sum(is_nan) > 0: # interpolation needed for filter_high_freq
+ angles[camn][label][is_nan] = np.interp(x(is_nan), x(~is_nan), angles[camn][label][~is_nan])
+
+ w = cutoff_frequency / (frame_rate / 2) # Normalize the frequency
+ b, a = signal.butter(2, w, 'low')
+
+ mean_angles[camn] += signal.filtfilt(b, a, angles[camn][label])
+
+ mean_angles[camn] = mean_angles[camn] / len(label_names_right)
+
+ min_angle = 0
+ for label_right in label_names_right:
+ label = label_right.replace('right', '')
+ in_out = np.where(abs(angles[camn][label] - mean_angles[camn]) > np.pi / 2)
+ angles_out[camn][label] = [s if i in in_out[0] else np.nan for i, s in enumerate(angles[camn][label])]
+ min_angle = np.min([np.min(angles[camn][label]), min_angle])
+
+ if show_plot:
+ fig1, axs1 = plt.subplots(nb_cam, 1)
+ fig1.suptitle('Angle of the line btw right and left wing part - Before unscrambling')
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ axs1[i].clear()
+ for label_right in label_names_right:
+ label = label_right.replace('right', '')
+ frames = list(range(0, len(angles_out[camn][label])))
+
+ axs1[i].scatter(frames, angles[camn][label], label=label, marker='.')
+ axs1[i].scatter(frames, angles_out[camn][label], c='r', label='out', marker='*')
+
+ axs1[i].plot(mean_angles[camn], 'k', label='mean')
+ axs1[i].set_ylabel('angle (rad)')
+ axs1[i].set_xlabel('frames - ' + label_right)
+
+ for i, frame in enumerate(skeleton2d[obj_name]['frames']):
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+ label = label_right.replace('right', '')
+
+ #if all([np.isnan(angles_out[camn][label][i]) for camn in range(1, nb_cam + 1)]): continue
+
+ xs_coord, ys_coord, frames_coord, likelihood_coords = [], [], [], []
+ for camn in range(1, nb_cam + 1):
+ if not np.isnan(angles_out[camn][label][i]): # Switch side
+ skeleton2d[obj_name][label_right][camn]['x'][i], skeleton2d[obj_name][label_left][camn]['x'][i] = \
+ skeleton2d[obj_name][label_left][camn]['x'][i], skeleton2d[obj_name][label_right][camn]['x'][i]
+ skeleton2d[obj_name][label_right][camn]['y'][i], skeleton2d[obj_name][label_left][camn]['y'][i] = \
+ skeleton2d[obj_name][label_left][camn]['y'][i], skeleton2d[obj_name][label_right][camn]['y'][i]
+ skeleton2d[obj_name][label_right][camn]['likelihood'][i], skeleton2d[obj_name][label_left][camn]['likelihood'][i] = \
+ skeleton2d[obj_name][label_left][camn]['likelihood'][i], skeleton2d[obj_name][label_right][camn]['likelihood'][i]
+
+ x_right = skeleton2d[obj_name][label_right][camn]['x'][i].copy()
+ y_right = skeleton2d[obj_name][label_right][camn]['y'][i].copy()
+ x_left = skeleton2d[obj_name][label_left][camn]['x'][i].copy()
+ y_left = skeleton2d[obj_name][label_left][camn]['y'][i].copy()
+
+ likelihood_right = skeleton2d[obj_name][label_right][camn]['likelihood'][i]
+ likelihood_left = skeleton2d[obj_name][label_left][camn]['likelihood'][i]
+
+ likelihood_coords.append([likelihood_right, likelihood_left])
+ xs_coord.append([x_right, x_left])
+ ys_coord.append([y_right, y_left])
+ frames_coord.append([frame, frame])
+
+ xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
+ reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False, sort_by_rmse=False)
+
+ # if len(xs_pose) == 0:
+ # xs_pose, ys_pose, zs_pose = [[np.nan, np.nan]], [[np.nan, np.nan]], [[np.nan, np.nan]]
+ # indexes = [[np.array([0, 0, 0]), np.array([1, 1, 1])]]
+ #
+ # else:
+ # rmses
+ # # order rmse
+ #
+ # cpt = 0
+ # while len(xs_pose[0]) > cpt:
+ # # Remove duplicate usage of points and give priority to first rows (with lowest rmse)
+ # # should have max two rows in the end (one per side)
+ #
+ # comparison_indexes = np.array([False] * nb_cam) # False if different from previous rows
+ # for j in range(1, cpt + 1):
+ # comparison_indexes += np.array(
+ # [indexes[0][cpt][k] == indexes[0][cpt - j][k] for k in range(0, nb_cam)])
+ #
+ # # Remove coordinates because it use at least one 2d points that was used already
+ # if comparison_indexes.any():
+ # xs_pose[0] = [s for k, s in enumerate(xs_pose[0]) if k != cpt]
+ # ys_pose[0] = [s for k, s in enumerate(ys_pose[0]) if k != cpt]
+ # indexes[0] = [s for k, s in enumerate(indexes[0]) if k != cpt]
+ # else:
+ # cpt += 1
+ #
+ # if len(xs_pose[0]) == 1:
+ # xs_pose[0].append(np.nan), ys_pose[0].append(np.nan), zs_pose[0].append(np.nan)
+ # indexes[0].append([abs(index - 1) for index in indexes[0]][0])
+ #
+ # # if np.nansum(indexes[0][0]) > np.nansum(indexes[0][1]) \
+ # # and indexes[0][1][1] == 0: # use top cam (#2) as reference for identifying side
+ # if indexes[0][1][1] == 0: # use top cam (#2) as reference for identifying side
+ # indexes[0][0], indexes[0][1] = indexes[0][1], indexes[0][0]
+ # xs_pose[0].reverse(), ys_pose[0].reverse(), zs_pose[0].reverse()
+
+ if len(xs_pose) == 0:
+ xs_pose, ys_pose, zs_pose = [[np.nan, np.nan]], [[np.nan, np.nan]], [[np.nan, np.nan]]
+ indexes = [[np.array([0, 0, 0]), np.array([1, 1, 1])]]
+
+ elif len(xs_pose[0]) == 1:
+ xs_pose[0].append(np.nan), ys_pose[0].append(np.nan), zs_pose[0].append(np.nan)
+ indexes[0].append([abs(index - 1) for index in indexes[0]][0])
+ #
+ # elif len(xs_pose[0]) > 2:
+ # indexes_top_right = np.array([s for s in indexes[0] if s[top_camn - 1] == 0.0]).astype(int)
+ # indexes_top_left = np.array([s for s in indexes[0] if s[top_camn - 1] == 1.0]).astype(int)
+ # rmses_top_right = np.array([rmses[0][i] for i, s in enumerate(indexes[0]) if s[top_camn - 1] == 0.0])
+ # rmses_top_left = np.array([rmses[0][i] for i, s in enumerate(indexes[0]) if s[top_camn - 1] == 1.0])
+ #
+ # indexes = [[np.array([0, 0, 0]), np.array([1, 1, 1])]]
+ # xs_pose_new, ys_pose_new, zs_pose_new = [[np.nan, np.nan]], [[np.nan, np.nan]], [[np.nan, np.nan]]
+ # if len(rmses_top_right) > 1 and len(rmses_top_left) > 1:
+ # combi = list(itertools.product(list(range(0, len(rmses_top_right))), list(range(0, len(rmses_top_left)))))
+ #
+ # new_combi = []
+ # for j in range(0, len(combi)): # Keep only the combinations that are compatible
+ # if (indexes_top_right[combi[j][0]] + indexes_top_left[combi[j][1]] == [1, 1, 1]).all():
+ # new_combi.append(combi[j])
+ #
+ # if not len(new_combi) == 0:
+ # rmse_combi = [np.sqrt(rmses_top_right[s[0]] ** 2 + rmses_top_left[s[1]] ** 2) for s in new_combi]
+ # best_combi = new_combi[np.argsort(rmse_combi)[0]]
+ #
+ # indexes = [[indexes_top_right[best_combi[0]], indexes_top_left[best_combi[1]]]]
+ #
+ # else:
+ # in_sort_right = np.argsort(rmses_top_right)
+ # in_sort_left = np.argsort(rmses_top_left)
+ #
+ # if rmses_top_right[in_sort_right[0]] < rmses_top_left[in_sort_left[0]]:
+ # indexes = [[indexes_top_right[in_sort_right[0]], abs(indexes_top_right[in_sort_right[0]] -1)]]
+ # else:
+ # indexes = [[abs(indexes_top_left[in_sort_left[0]] -1), indexes_top_left[in_sort_left[0]]]]
+ #
+ # xs_pose_new[0][0], ys_pose_new[0][0], zs_pose_new[0][0] = \
+ # xs_pose[0][indexes[0][0][0]], ys_pose[0][indexes[0][0][1]], zs_pose[0][indexes[0][0][2]]
+ # xs_pose_new[0][1], ys_pose_new[0][1], zs_pose_new[0][1] = \
+ # xs_pose[0][indexes[0][1][0]], ys_pose[0][indexes[0][1][1]], zs_pose[0][indexes[0][1][2]]
+ #
+ # elif not len(rmses_top_right) == 0 and len(rmses_top_left) == 0:
+ # in_sort_right = np.argsort(rmses_top_right)
+ # in_sort_left = np.argsort(rmses_top_left)
+ #
+ # if len(rmses_top_right) == 0:
+ # indexes[0][0] = abs(indexes_top_left[in_sort_left[0]] - 1)
+ # else:
+ # indexes[0][0] = indexes_top_right[in_sort_right[0]]
+ # xs_pose_new[0][0], ys_pose_new[0][0], zs_pose_new[0][0] = \
+ # xs_pose[0][indexes[0][0][0]], ys_pose[0][indexes[0][0][1]], zs_pose[0][indexes[0][0][2]]
+ #
+ # if len(rmses_top_left) == 0:
+ # indexes[0][1] = abs(indexes_top_right[in_sort_right[0]] - 1)
+ # else:
+ # indexes[0][1] = indexes_top_left[in_sort_left[0]]
+ # xs_pose_new[0][1], ys_pose_new[0][1], zs_pose_new[0][1] = \
+ # xs_pose[0][indexes[0][1][0]], ys_pose[0][indexes[0][1][1]], zs_pose[0][indexes[0][1][2]]
+ #
+ # xs_pose, ys_pose, zs_pose = xs_pose_new, ys_pose_new, zs_pose_new
+ #
+ # if indexes[0][1][top_camn-1] == 0: # use top cam as reference for identifying side
+ # indexes[0][0], indexes[0][1] = indexes[0][1], indexes[0][0]
+ # xs_pose[0].reverse(), ys_pose[0].reverse(), zs_pose[0].reverse()
+ #
+ # if (indexes[0] != np.array([0, 0, 0])).any():
+ # for j, camn in enumerate(range(1, nb_cam + 1)):
+ # if not np.isnan(indexes[0][0][j]) and not np.isnan(indexes[0][1][j]):
+ # in_right = int(indexes[0][0][j])
+ # skeleton2d[obj_name][label_right][camn]['x'][i] = xs_coord[j][in_right]
+ # skeleton2d[obj_name][label_right][camn]['y'][i] = ys_coord[j][in_right]
+ # skeleton2d[obj_name][label_right][camn]['likelihood'][i] = likelihood_coords[j][in_right]
+ #
+ # in_left = int(indexes[0][1][j])
+ # skeleton2d[obj_name][label_left][camn]['x'][i] = xs_coord[j][in_left]
+ # skeleton2d[obj_name][label_left][camn]['y'][i] = ys_coord[j][in_left]
+ # skeleton2d[obj_name][label_left][camn]['likelihood'][i] = likelihood_coords[j][in_left]
+
+ skeleton3d[obj_name][label_right]['x'][i], skeleton3d[obj_name][label_right]['y'][i], \
+ skeleton3d[obj_name][label_right]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
+
+ skeleton3d[obj_name][label_left]['x'][i], skeleton3d[obj_name][label_left]['y'][i], \
+ skeleton3d[obj_name][label_left]['z'][i] = xs_pose[0][1], ys_pose[0][1], zs_pose[0][1]
+
+ for label in other_label_names:
+ xs_coord, ys_coord, frames_coord = [], [], []
+ for camn in range(1, nb_cam + 1):
+ xs_coord.append([skeleton2d[obj_name][label][camn]['x'][i].copy()])
+ ys_coord.append([skeleton2d[obj_name][label][camn]['y'][i].copy()])
+ frames_coord.append([frame])
+
+ xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
+ reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
+
+ if len(xs_pose) == 0:
+ skeleton3d[obj_name][label]['x'][i], skeleton3d[obj_name][label]['y'][i], \
+ skeleton3d[obj_name][label]['z'][i] = np.nan, np.nan, np.nan
+ else:
+ skeleton3d[obj_name][label]['x'][i], skeleton3d[obj_name][label]['y'][i], \
+ skeleton3d[obj_name][label]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
+
+ if show_plot:
+ fig2, axs2 = plt.subplots(nb_cam, 1)
+ fig2.suptitle('Angle of the line btw right and left wing part - After unscrambling')
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ axs2[i].clear()
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+ x_right = skeleton2d[obj_name][label_right][camn]['x']
+ y_right = skeleton2d[obj_name][label_right][camn]['y']
+ x_left = skeleton2d[obj_name][label_left][camn]['x']
+ y_left = skeleton2d[obj_name][label_left][camn]['y']
+
+ angles = np.arctan2((y_right - y_left), (x_right - x_left))
+ #angles = np.arctan((y_right - y_left) / (x_right - x_left))
+ #angles[~np.isnan(angles)] = np.unwrap(angles[~np.isnan(angles)] * 2, discont=2 * np.pi * 0.8) / 2
+ angles = np.mod(angles, 2*np.pi)
+
+ axs2[i].plot(angles, label=label_right.replace('right', ''))
+ axs2[i].set_ylabel('angle (rad)')
+ axs2[i].set_xlabel('frames - ' + label_right)
+ plt.legend
+ plt.show()
+
+
+def reverse_processes_2d_points(skeleton2d, process_dict, threshold_likelihood, img_size, save_path, show_plot=False):
+
+ nb_cam = len(process_dict.keys())
+
+ [width_img, height_img] = img_size
+ for camn in range(1, nb_cam + 1):
+ process_names = [process_dict[camn][process_num]['fn'] for process_num in process_dict[camn].keys()]
+
+ for obj_name in skeleton2d['obj_names']:
+ if 'stitch' in process_names or 'rotate' in process_names:
+ if 'crop' in process_names:
+ crop_num = process_names.index('crop')
+ crop_num = crop_num[-1] if isinstance(crop_num, list) else crop_num
+ width_crop = process_dict[camn][crop_num]['kwargs']['width_crop']
+ height_crop = process_dict[camn][crop_num]['kwargs']['height_crop']
+
+ else:
+ width_crop, height_crop = width_img, height_img
+
+ if 'crop' in process_names:
+ crop_paths = os.path.join(save_path, '_Cropped')
+ crop_path = crop_paths[process_dict[camn][0]['camn']]
+ csv_path = glob.glob(os.path.join(crop_path, process_dict[camn][0]['rec_names'][camn],
+ '*' + obj_name + '-2d_points.csv'))
+
+ pts_csv = np.genfromtxt(csv_path[0], delimiter=',', skip_header=0, names=True)
+
+ temp_set = set(skeleton2d[obj_name]['frames'])
+ in_frame = [i for i, val in enumerate(pts_csv['frame']) if val in temp_set]
+
+ width_crop = process_dict[camn][crop_num]['kwargs']['width_crop']
+ height_crop = process_dict[camn][crop_num]['kwargs']['height_crop']
+
+ x_crop = pts_csv['x_px'][in_frame] - width_crop / 2
+ y_crop = pts_csv['y_px'][in_frame] - height_crop / 2
+
+ for i in range(0, len(in_frame)):
+ x_crop[i] = np.max([np.min([x_crop[i], width_img - width_crop]), 0])
+ y_crop[i] = np.max([np.min([y_crop[i], height_img - height_crop]), 0])
+
+ for label in skeleton2d[obj_name]['label_names']:
+ if 'stitch' in process_names:
+ skeleton2d[obj_name][label][camn]['x'] += - width_crop * (camn - 1)
+
+ if 'rotate' in process_names and \
+ camn in process_dict[camn][process_names.index('rotate')]['kwargs']['camn_to_process']:
+ x = skeleton2d[obj_name][label][camn]['x'].copy() - width_crop / 2
+ y = skeleton2d[obj_name][label][camn]['y'].copy() - height_crop / 2
+ teta = np.deg2rad(process_dict[camn][process_names.index('rotate')]['kwargs']['degrees'])
+
+ skeleton2d[obj_name][label][camn]['x'] = x * np.cos(teta) - y * np.sin(teta) + width_crop / 2
+ skeleton2d[obj_name][label][camn]['y'] = x * np.sin(teta) + y * np.cos(teta) + height_crop / 2
+
+ if 'crop' in process_names:
+ skeleton2d[obj_name][label][camn]['x'] += x_crop
+ skeleton2d[obj_name][label][camn]['y'] += y_crop
+
+ if show_plot:
+ raw_rec_path = os.path.join(process_dict[camn][0]['paths'][process_dict[camn][0]['camn']],
+ process_dict[camn][0]['rec_names'][process_dict[camn][0]['camn']])
+
+ all_files = sorted(os.listdir(raw_rec_path), key=lambda s: s[s.rfind('.') - nb_leading_zeros:s.rfind('.')])
+ all_images = [s for s in all_files if '.{0}'.format(image_format) in s]
+
+ fig = plt.figure()
+ for i, image in enumerate(all_images):
+ if i % 500: continue # plot every 100 frames
+ ax = fig.add_subplot(111)
+ plt.clf()
+ plt.imshow(Image.open(os.path.join(raw_rec_path, image)))
+ # ax.scatter(x_crop + width_crop/2, y_crop + height_crop/2, marker='o')
+ for label in skeleton2d[obj_name]['label_names']:
+ # if 'wing' not in label: continue
+ if skeleton2d[obj_name][label][camn]['likelihood'][i] > threshold_likelihood:
+ ax.scatter(skeleton2d[obj_name][label][camn]['x'][i],
+ skeleton2d[obj_name][label][camn]['y'][i], 1)
+ ax.set_xlabel('x (m)')
+ ax.set_ylabel('y (m)')
+ plt.pause(0.1)
+ # plt.show()
+
+
+def test_unwrap(skeleton2d, dlt_path, reconstructor_settings: Reconstructor3DSettings):
+ reconstructor = Reconstructor3D.read_dlt_coefs(dlt_path)
+ reconstructor.settings = reconstructor_settings
+ nb_cam = reconstructor.nb_cam
+
+ # Unscamble case where side is wrong on only a minority of cam views
+ skeleton3d = {'obj_names': skeleton2d['obj_names'], 'model_name': skeleton2d['model_name']}
+ for obj_name in skeleton3d['obj_names']:
+ label_names_right = [s for s in skeleton2d[obj_name]['label_names'] if 'right' in s]
+
+ skeleton3d[obj_name] = {'frames': skeleton2d[obj_name]['frames'],
+ 'label_names': skeleton2d[obj_name]['label_names']}
+ for label in skeleton3d[obj_name]['label_names']:
+ skeleton3d[obj_name][label] = {'x': np.nan * np.ones(np.size(skeleton3d[obj_name]['frames'])),
+ 'y': np.nan * np.ones(np.size(skeleton3d[obj_name]['frames'])),
+ 'z': np.nan * np.ones(np.size(skeleton3d[obj_name]['frames']))}
+
+ for i, frame in enumerate(skeleton3d[obj_name]['frames']):
+ for label in skeleton3d[obj_name]['label_names']:
+ xs_coord, ys_coord, frames_coord = [], [], []
+ for camn in range(1, nb_cam + 1):
+ xs_coord.append([skeleton2d[obj_name][label][camn]['x'][i].copy()])
+ ys_coord.append([skeleton2d[obj_name][label][camn]['y'][i].copy()])
+ frames_coord.append([frame])
+
+ xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
+ reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
+
+ if len(xs_pose) == 0:
+ skeleton3d[obj_name][label]['x'][i], skeleton3d[obj_name][label]['y'][i], \
+ skeleton3d[obj_name][label]['z'][i] = np.nan, np.nan, np.nan
+ else:
+ skeleton3d[obj_name][label]['x'][i], skeleton3d[obj_name][label]['y'][i], \
+ skeleton3d[obj_name][label]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
+
+ fig1, axs1 = plt.subplots(nb_cam, 1)
+ for i, camn in enumerate(range(1, nb_cam + 1)):
+ axs1[i].clear()
+ for label_right in label_names_right:
+ label_left = label_right.replace('right', 'left')
+ x_right = skeleton2d[obj_name][label_right][camn]['x']
+ y_right = skeleton2d[obj_name][label_right][camn]['y']
+ x_left = skeleton2d[obj_name][label_left][camn]['x']
+ y_left = skeleton2d[obj_name][label_left][camn]['y']
+
+ angle = np.arctan((y_right - y_left) / (x_right - x_left))
+ angle[~np.isnan(angle)] = np.unwrap(angle[~np.isnan(angle)] * 2, discont=2 * np.pi * 0.8) / 2
+
+ angle2 = np.arctan2((y_right - y_left), (x_right - x_left))
+ median_diff_angle = np.nanmedian(angle2 - angle)
+ if abs(abs(median_diff_angle) - np.pi) < 0.2:
+ if median_diff_angle > 0:
+ angle2 -= np.pi
+ elif median_diff_angle < 0:
+ angle2 += np.pi
+
+ in_diff = np.array([False] * len(angle))
+ is_not_nan = ~np.isnan(angle) & ~np.isnan(angle2)
+ in_diff[is_not_nan] = abs(angle2[is_not_nan] - angle[is_not_nan]) > np.pi / 2
+
+ axs1[i].plot(skeleton2d[obj_name]['frames'], angle) # angle of the line btw the two sides
+ axs1[i].plot(np.array(skeleton2d[obj_name]['frames'])[in_diff], angle2[in_diff], marker='*',
+ linestyle='None')
+
+ axs1[i].set_ylabel('angle btw right and left (rad)')
+ axs1[i].set_xlabel('frames - ' + label_right)
+
+ plt.show()
+
+
+def recon3d_dlc(skeleton2d, dlt_path, reconstructor_settings: Reconstructor3DSettings):
+ reconstructor = Reconstructor3D.read_dlt_coefs(dlt_path)
+ reconstructor.settings = reconstructor_settings
+
+ # Unscamble case where side is wrong on only a minority of cam views
+ skeleton3d = {'obj_names': skeleton2d['obj_names'], 'model_name': skeleton2d['model_name']}
+ for obj_name in skeleton3d['obj_names']:
+ # label_names_right = [s for s in skeleton2d[obj_name]['label_names'] if 'right' in s]
+ # other_label_names = [s for s in skeleton2d[obj_name]['label_names'] if
+ # 'right' not in s or 'left' not in s]
+
+ skeleton3d[obj_name] = {'frames': skeleton2d[obj_name]['frames'],
+ 'label_names': skeleton2d[obj_name]['label_names']}
+ for label in skeleton3d[obj_name]['label_names']:
+ skeleton3d[obj_name][label] = {'x': np.nan * np.ones(np.size(skeleton2d[obj_name]['frames'])),
+ 'y': np.nan * np.ones(np.size(skeleton2d[obj_name]['frames'])),
+ 'z': np.nan * np.ones(np.size(skeleton2d[obj_name]['frames']))}
+
+ for i, frame in enumerate(skeleton2d[obj_name]['frames']):
+ for label in skeleton3d[obj_name]['label_names']:
+ xs_coord, ys_coord, frames_coord = [], [], []
+
+ nb_cam = len(skeleton2d[obj_name][label].keys())
+ for camn in range(1, nb_cam + 1):
+ xs_coord.append([skeleton2d[obj_name][label][camn]['x'][i].copy()])
+ ys_coord.append([skeleton2d[obj_name][label][camn]['y'][i].copy()])
+ frames_coord.append([frame])
+
+ xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
+ reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
+
+ if len(xs_pose) == 0:
+ skeleton3d[obj_name][label]['x'][i], skeleton3d[obj_name][label]['y'][i], \
+ skeleton3d[obj_name][label]['z'][i] = np.nan, np.nan, np.nan
+ else:
+ skeleton3d[obj_name][label]['x'][i], skeleton3d[obj_name][label]['y'][i], \
+ skeleton3d[obj_name][label]['z'][i] = xs_pose[0][0], ys_pose[0][0], zs_pose[0][0]
\ No newline at end of file
diff --git a/process/images_preprocessing.py b/process/images_preprocessing.py
new file mode 100644
index 0000000000000000000000000000000000000000..87dbad4e0093b75f07a92c1c797b7acb6f18bbb3
--- /dev/null
+++ b/process/images_preprocessing.py
@@ -0,0 +1,158 @@
+import os
+import cv2
+
+import numpy as np
+from PIL import Image
+
+from typing import List
+
+import images.utils as img_utils
+from images.read import read_image
+
+from images.utils import find_common_substrings
+
+
+def crop_image(img, coords, height, width):
+
+ x_pxs, y_pxs = np.array([]), np.array([])
+ for x_name in [s for s in coords.keys() if 'x_' in s]:
+ y_name = x_name.replace('x_', 'y_')
+
+ x_pxs = np.append(x_pxs, coords[x_name])
+ y_pxs = np.append(y_pxs, coords[y_name])
+
+ assert x_pxs.size and y_pxs.size
+
+ return img_utils.crop(img, x_pxs, y_pxs, height, width)
+
+
+def crop_all_images(coords_objs, image_names, image_path, image_save_path, height, width):
+
+ obj_names = coords_objs.keys()
+
+ for ind_img, image_name in enumerate(image_names):
+
+ img = read_image(os.path.join(image_path, image_name))
+
+ img_cropped = []
+ for obj_name in obj_names:
+ coords_obj = {key: coords_objs[obj_name][key][ind_img] for key in coords_objs[obj_name].keys()}
+
+ img_cropped.append(crop_image(img, coords_obj, height, width))
+
+ image_save_name = image_name[:image_name.rfind('.')] + '-' + obj_name + image_name[image_name.rfind('.'):]
+ img_cropped[-1].save(os.path.join(image_save_path, image_save_name))
+ img_cropped[-1].close()
+
+ img.close()
+
+
+def stitch_all_images(image_names, image_paths, frames, image_save_paths, main_camn):
+ # TODO make this function work with an MultiImageSequence (size: nb_cam*nb_frame)
+ # or list of images (size: nb_cam), not an image_path, image_format ...
+ # + use this function in multiprocesses in BatchProcessing
+
+ nb_cam = len(image_names.keys())
+
+ frames_set = set()
+ for camn in range(1, nb_cam + 1):
+ frames_set = frames_set & set(frames[camn]) if len(frames_set) > 0 else set(frames[camn])
+
+ frames_set = np.sort(list(frames_set))
+
+ index_i = {}
+ for camn in range(1, nb_cam + 1):
+ index_i[camn] = [frames[camn].index(frame) for frame in frames_set]
+
+ if not os.path.exists(image_save_paths[main_camn]): os.makedirs(image_save_paths[main_camn])
+
+ for i, _ in enumerate(index_i[main_camn]):
+ imgs = {}
+ for camn in range(1, nb_cam + 1):
+ imgs[camn] = Image.open(os.path.join(image_paths[camn], image_names[camn][index_i[camn][i]]))
+
+ img = img_utils.stitch_images(imgs, nb_cam)
+
+ if image_names[main_camn][index_i[main_camn][i]].find('.jpg') != -1 and img.mode != 'RGB': img = img.convert('RGB')
+
+ img.save(os.path.join(image_save_paths[main_camn], image_names[main_camn][index_i[main_camn][i]]), compression='lzw')
+
+
+def stitch_images_multiprocessing(img, image_names, image_rec_paths, frames, image_save_paths, obj_names, multi_processes_num,
+ multi_processes_nums, main_camn):
+
+ # TODO Merge this function with stitch_images, make them work on a MultiImageSequence
+
+ nb_cam = len(image_names.keys())
+
+ frames_set = set()
+ for camn in range(1, nb_cam + 1):
+ frames_set = frames_set & set(frames[camn]) if len(frames_set) > 0 else set(frames[camn])
+
+ frames_set = np.sort(list(frames_set))
+
+ index_i = {}
+ for camn in range(1, nb_cam + 1):
+ index_i[camn] = [frames[camn].index(frame) for frame in frames_set]
+
+ for obj_name in obj_names:
+ # if obj_name == 'obj0' and 'from_fn_name' in kwargs.keys(): break
+
+ for camn in range(1, nb_cam + 1):
+ if obj_name not in img[camn].keys():
+ img[camn][obj_name] = {}
+
+ for i, _ in enumerate(index_i[main_camn]):
+ imgs = {}
+ for camn in range(1, nb_cam + 1):
+ if i not in img[camn][obj_name].keys():
+ img[camn][obj_name][i] = read_image(os.path.join(image_rec_paths[camn], image_names[camn][i]))
+
+ imgs[camn] = img[camn][obj_name][i]
+ img[camn][obj_name][i] = []
+
+ img[main_camn][obj_name][i] = img_utils.stitch_images(imgs, nb_cam)
+
+ if multi_processes_num == multi_processes_nums[-1]:
+ for i, image_name in enumerate(image_names[camn]):
+ if 'obj0' == obj_name:
+ image_save_name = image_name
+ else:
+ image_save_name = image_name[:image_name.rfind('.')] + '-' + obj_name + image_name[image_name.rfind('.'):]
+
+ if image_name.find('.jpg') != -1 and img[main_camn][obj_name][i].mode != 'RGB':
+ img[main_camn][obj_name][i] = img[main_camn][obj_name][i].convert('RGB')
+
+ img[main_camn][obj_name][i].save(os.path.join(image_save_paths[main_camn], image_save_name))
+ img[main_camn][obj_name][i].close()
+
+ return img
+
+
+def save_avi(all_image_names: List[str], image_path: str, rec_name: str, save_path: str, frame_rate: int = 24,
+ step_frame: int = 1, lossless: bool = False) -> None:
+
+ obj_names = np.unique([s[s.rfind('-obj')+1:s.rfind('.')] for s in all_image_names])
+
+ for obj_name in obj_names:
+ image_names = sorted([i for i in all_image_names if obj_name in i], key=lambda s: s[len(rec_name):s.rfind('.')])
+ save_name = rec_name + '-' + obj_name + '.avi'
+
+ img = cv2.imread(os.path.join(image_path, image_names[0]), cv2.IMREAD_UNCHANGED)
+ is_rgb = len(img.shape) == 3
+ if is_rgb: height_img, width_img, _ = img.shape
+ else: height_img, width_img = img.shape
+
+ # commun: 'MP42', 'XVID' or lossless: 'HFYU'
+ codec = cv2.VideoWriter_fourcc(*'HFYU') if lossless else cv2.VideoWriter_fourcc(*'MP42')
+ writer = cv2.VideoWriter(os.path.join(save_path, save_name), codec, frame_rate, (width_img, height_img,))
+
+ for i in np.arange(0, len(image_names), step_frame):
+ img_16bit = cv2.imread(os.path.join(image_path, image_names[i]), cv2.IMREAD_UNCHANGED)
+ if is_rgb: writer.write(img_16bit)
+ else: writer.write(cv2.cvtColor(img_16bit, cv2.COLOR_GRAY2BGR))
+ #cv2.imshow(obj_name, img_16bit)
+ #cv2.waitKey(1)
+
+ writer.release()
+
diff --git a/process/point_tracking.py b/process/point_tracking.py
new file mode 100644
index 0000000000000000000000000000000000000000..d0a73c834a06393e26f97c42d22ef6fb8a64c8e1
--- /dev/null
+++ b/process/point_tracking.py
@@ -0,0 +1,155 @@
+import glob
+import os
+
+import numpy as np
+import matplotlib.pyplot as plt
+from PIL import Image
+
+from images.imagesequence import ImageSequence
+
+from point_tracker.reconstructor3d import Reconstructor3D
+from point_tracker.tracker2d import Tracker2D, BlobDetectorSettings, BackgroundSubtractorSettings
+from point_tracker.reconstructor3d import Reconstructor3DSettings
+
+
+def track_2d_objects(image_sequence: ImageSequence, save_name: str, save_path: str,
+ back_subtractor_settings: BackgroundSubtractorSettings,
+ blob_detector_settings: BlobDetectorSettings, show_plot: bool = False):
+
+ image_names = image_sequence.get_names()
+ image_path = image_sequence.get_path()
+
+ frames = image_sequence.get_numbers()
+
+ tracker = Tracker2D()
+ tracker.get_back_subtractor(back_subtractor_settings)
+ tracker.get_blob_detector(blob_detector_settings)
+
+ tracker.load_images(image_names, image_path, frames)
+
+ tracker.do_tracking()
+
+ if show_plot:
+ fig = plt.figure()
+ ax = fig.add_subplot(111)
+ plt.imshow(Image.open(os.path.join(image_path, image_names[0])))
+ ax.scatter(tracker.points['x'], tracker.points['y'])
+ ax.set_xlabel('x (m)')
+ ax.set_ylabel('y (m)')
+ plt.show()
+
+ tracker.save_csv(save_name, save_path)
+
+
+def interp_2d_coords(frames, csv_name, csv_path, save_path, image_size, show_plot=False):
+ width_img, height_img = image_size
+
+ old_obj_dict = np.genfromtxt(os.path.join(csv_path, csv_name), delimiter=',', skip_header=0, names=True)
+
+ # is_not_nan = ~np.isnan(old_obj_dict['x_px']) & ~np.isnan(old_obj_dict['y_px'])
+ # old_obj_dict = old_obj_dict[is_not_nan]
+
+ x_px = np.interp(frames, old_obj_dict['frame'], old_obj_dict['x_px'])
+ y_px = np.interp(frames, old_obj_dict['frame'], old_obj_dict['y_px'])
+
+ for i, _ in enumerate(frames):
+ if x_px[i] > width_img: x_px[i] = width_img
+ if y_px[i] > height_img: y_px[i] = height_img
+ if x_px[i] < 0: x_px[i] = 0
+ if y_px[i] < 0: y_px[i] = 0
+
+ np.savetxt(os.path.join(save_path, csv_name), np.c_[frames, x_px, y_px], delimiter=',', header='frame,x_px,y_px')
+ obj_dict = {'frame': frames, 'x_px': x_px, 'y_px': y_px}
+
+ if show_plot:
+ fig = plt.figure()
+ ax = fig.add_subplot(111)
+ ax.scatter(obj_dict['x_px'], obj_dict['y_px'], marker='*')
+ ax.scatter(old_obj_dict['x_px'], old_obj_dict['y_px'], marker='x')
+ ax.set_xlabel('x (m)')
+ ax.set_ylabel('y (m)')
+ plt.xlim(0, width_img)
+ plt.ylim(height_img, 0)
+ plt.show()
+
+ return obj_dict
+
+
+def interp_2d_coords_of_objs(frames, csv_path, save_path, image_size, show_plot=False):
+
+ csv_names = glob.glob(os.path.join(csv_path, '*-obj*-2d_points.csv'))
+ csv_names = [os.path.basename(os.path.normpath(csv_name)) for csv_name in csv_names]
+
+ obj_names = [csv_name[csv_name.rfind('-obj') + 1:csv_name.find('-2d_points.csv')] for csv_name in csv_names]
+
+ coords_objs = {}
+ for i, csv_name in enumerate(csv_names):
+ if len(obj_names) == 0:
+ coords_objs[obj_names[i]] = {}
+
+ else:
+ coords_objs[obj_names[i]] = \
+ interp_2d_coords(frames, csv_name, csv_path, save_path, image_size, show_plot)
+
+ return coords_objs
+
+
+def reconstruct_3d_points(pts_dict, dlt_path, save_names, save_paths,
+ reconstructor_settings: Reconstructor3DSettings, save_csv: bool = True):
+
+ reconstructor = Reconstructor3D.read_dlt_coefs(dlt_path)
+ reconstructor.settings = reconstructor_settings
+
+ nb_cam = reconstructor.nb_cam
+
+ xs_coord = np.array([pts_dict[camn]['x_px'] for camn in range(1, nb_cam + 1)])
+ ys_coord = np.array([pts_dict[camn]['y_px'] for camn in range(1, nb_cam + 1)])
+ frames_coord = np.array([pts_dict[camn]['frame'] for camn in range(1, nb_cam + 1)])
+
+ xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses, = \
+ reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
+
+ save_names_list = [save_names[camn] for camn in range(1, nb_cam + 1)]
+ save_paths_list = [save_paths[camn] for camn in range(1, nb_cam + 1)]
+ if save_csv: reconstructor.save_points_csv(save_names_list, save_paths_list, save_type='all_frames')
+
+ return xs_pose, ys_pose, zs_pose, frames_pose, indexes, rmses
+
+
+def reconstruct_3d_tracks(pts_dict, dlt_path, save_names, save_paths, reconstructor_settings: Reconstructor3DSettings,
+ show_plot=False, save_csv=True):
+
+ reconstructor = Reconstructor3D.read_dlt_coefs(dlt_path)
+ reconstructor.settings = reconstructor_settings
+ nb_cam = reconstructor.nb_cam
+
+ xs_coord = np.array([pts_dict[camn]['x_px'] for camn in range(1, nb_cam + 1)])
+ ys_coord = np.array([pts_dict[camn]['y_px'] for camn in range(1, nb_cam + 1)])
+ frames_coord = np.array([pts_dict[camn]['frame'] for camn in range(1, nb_cam + 1)])
+
+ xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts, _, = \
+ reconstructor.recon_objs(xs_coord, ys_coord, frames_coord, with_missing=False)
+
+ tracks_dict = reconstructor.recon_tracks(xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts)
+
+ if not xs_pose or not tracks_dict['obj']: return
+
+ save_names_list = [save_names[camn] for camn in range(1, nb_cam + 1)]
+ save_paths_list = [save_paths[camn] for camn in range(1, nb_cam + 1)]
+ if save_csv: reconstructor.save_tracks_csv(save_names_list, save_paths_list)
+
+ if show_plot:
+ xs_pose, ys_pose, zs_pose = np.concatenate(xs_pose), np.concatenate(ys_pose), np.concatenate(zs_pose)
+
+ fig = plt.figure()
+ ax = fig.add_subplot(111, projection='3d')
+ ax.scatter(xs_pose, ys_pose, zs_pose, marker='+')
+ for nb_obj in tracks_dict['obj'].keys():
+ ax.scatter(tracks_dict['obj'][nb_obj]['x'], tracks_dict['obj'][nb_obj]['y'],
+ tracks_dict['obj'][nb_obj]['z'], marker='o')
+ ax.set_xlabel('x (m)')
+ ax.set_ylabel('y (m)')
+ ax.set_zlabel('z (m)')
+ plt.show()
+
+ return tracks_dict
diff --git a/process/skeleton_fitting.py b/process/skeleton_fitting.py
new file mode 100644
index 0000000000000000000000000000000000000000..1030d189481354830daee0d2381b6953a26cfbd6
--- /dev/null
+++ b/process/skeleton_fitting.py
@@ -0,0 +1,1431 @@
+import copy, glob, os, time
+from functools import partial
+from multiprocessing import Pool
+
+import numpy as np
+from PIL import Image, ImageDraw
+from matplotlib import pyplot as plt
+from scipy import signal, optimize
+from sympy import symbols, solve
+
+from camera import calib
+from images.read import read_image
+from process.smoothing_filtering import filter_interp_params
+from skeleton_fitter.utils import length_from_vector, reproject_skeleton3d_to2d, angle_from_vectors
+from skeleton_fitter.optimiser import optim_fit_body_params, optim_fit_limbs_params
+
+# TODO remove dependencies to insect body/wing/hybrid functions
+# use: import importlib
+# animal_type = 'insect'
+# skeleton_type = ['body_slim', 'wings', 'hybrid_wings_body_slim']
+# module = importlib.import_module('skeleton_fitter.skeleton_modules.' + animal_type + '_' + skeleton_type)
+import skeleton_fitter.skeleton_modules.insect_body_slim as body
+import skeleton_fitter.skeleton_modules.insect_wings as wings
+import skeleton_fitter.skeleton_modules.insect_hybrid_wings_body_slim as hybrid
+
+
+def load_skeleton2d_from_csv(model_name, paths, main_camn):
+
+ nb_cam = len(paths.keys())
+
+ csv_paths = glob.glob(os.path.join(paths[main_camn], '*-2d_points.csv'))
+ obj_names = np.unique([csv_path[len(paths[main_camn]) + 1:csv_path.index('-')] for csv_path in csv_paths])
+
+ skeleton2d = {'obj_names': obj_names, 'model_name': model_name}
+ for obj_name in obj_names:
+ label_names = np.unique([csv_path[csv_path.index('-') + 1:csv_path.index('-2d_points.csv')] for csv_path in csv_paths])
+
+ skeleton2d[obj_name] = {'frames': [], 'label_names': label_names}
+ for label in label_names:
+ skeleton2d[obj_name][label] = {1: {}}
+ for camn in range(1, nb_cam + 1):
+ pts_csv = np.genfromtxt(os.path.join(paths[camn], obj_name + '-' + label + '-2d_points.csv'),
+ delimiter=',', skip_header=0, names=True)
+
+ skeleton2d[obj_name]['frames'] = \
+ np.unique(np.concatenate((skeleton2d[obj_name]['frames'], pts_csv['frame'])))
+ skeleton2d[obj_name][label][camn] = \
+ {'x': pts_csv['x_px'], 'y': pts_csv['y_px'], 'likelihood': pts_csv['likelihood']}
+
+ return skeleton2d
+
+
+def load_skeleton3d_from_csv(model_name, paths, main_camn):
+
+ nb_cam = len(paths.keys())
+
+ csv_paths = glob.glob(os.path.join(paths[main_camn], '*-3d_points.csv'))
+ obj_names = np.unique([csv_path[len(paths[main_camn]) + 1:csv_path.index('-')] for csv_path in csv_paths])
+
+ if len(csv_paths) == 0:
+ raise ValueError("Couldn't find any *-3d_points.csv file (path: {0})".format(os.path.join(paths[main_camn])))
+
+ skeleton3d = {'obj_names': obj_names, 'model_name': model_name}
+ for obj_name in obj_names:
+ obj_csv_paths = [csv_path for csv_path in csv_paths if obj_name in csv_path]
+ label_names = np.unique([csv_path[csv_path.index(obj_name + '-') + len(obj_name) +1:
+ csv_path.index('-3d_points.csv')] for csv_path in obj_csv_paths])
+
+ skeleton3d[obj_name] = {'frames': [], 'label_names': label_names}
+ for label in label_names:
+ skeleton3d[obj_name][label] = {1: {}}
+ for camn in range(1, nb_cam + 1):
+ pts_csv = np.genfromtxt(os.path.join(paths[camn], obj_name + '-' + label + '-3d_points.csv'),
+ delimiter=',', skip_header=0, names=True)
+
+ skeleton3d[obj_name]['frames'] = \
+ np.unique(np.concatenate((skeleton3d[obj_name]['frames'], pts_csv['frame'])))
+ skeleton3d[obj_name][label] = {'x': pts_csv['x'], 'y': pts_csv['y'], 'z': pts_csv['z']}
+
+ return skeleton3d
+
+
+def save_skeleton_params_in_csv(skeleton3d, wings_params_init, body_params, wings_params, save_path):
+ for obj_name in skeleton3d['obj_names']:
+ body_params_names = body_params[obj_name].keys()
+
+ params_list, params_names = [skeleton3d[obj_name]['frames']], 'frame'
+ for param_name in body_params_names:
+ params_list.append(body_params[obj_name][param_name])
+ params_names += ',' + param_name
+
+ for side in wings_params_init.keys():
+ for param_name in wings_params[obj_name][side].keys():
+ if param_name in body_params_names: continue
+ params_list.append(wings_params[obj_name][side][param_name])
+ params_names += ',' + param_name + '_' + side
+
+ np.savetxt(os.path.join(save_path, 'skeleton_parameters-' + obj_name + '.csv'),
+ np.c_[np.transpose(params_list)], delimiter=',', header=params_names)
+
+
+def save_obj_metric_in_csv(skeleton3d, wings_params_init, list_dict, list_name, save_path):
+ for obj_name in skeleton3d['obj_names']:
+ params_list, params_names = [skeleton3d[obj_name]['frames']], 'frame'
+ if isinstance(list_dict[obj_name], dict):
+ for side in wings_params_init.keys():
+ params_list.append(list_dict[obj_name][side])
+ params_names += ',' + list_name + '_' + side
+ else:
+ params_list.append(list_dict[obj_name])
+ params_names += ',' + list_name
+
+ np.savetxt(os.path.join(save_path, list_name + '-' + obj_name + '.csv'),
+ np.c_[np.transpose(params_list)], delimiter=',', header=params_names)
+
+
+class SkeletonFitting:
+ def __init__(self):
+ self.nb_process = 5
+
+ self.cutoff_frequency = 100 # Hz (for Butterworth filter on body params)
+
+ self.skeleton2d, self.skeleton3d = {}, {} # TODO should get these from arguments
+
+ # TODO remove following hybrid skeleton
+ self.hybrid_skeleton2d = copy.deepcopy(self.skeleton2d)
+ self.hybrid_skeleton3d = copy.deepcopy(self.skeleton3d)
+
+ # TODO Only use one module at a time (either body, wings or hybrid)
+ # + loop through module name using AnimalSettings class (from fly.py)
+ self.body_sets = body.ModuleSettings()
+ self.wings_sets = wings.ModuleSettings()
+ self.hybrid_sets = hybrid.ModuleSettings()
+
+ def load_params_from_csv(self, path):
+ obj_names = self.skeleton2d['obj_names']
+
+ body_params, wings_params = {}, {}
+ for obj_name in obj_names:
+ body_params[obj_name], wings_params[obj_name] = {}, {}
+ params_csv = np.genfromtxt(os.path.join(path, 'skeleton_parameters-' + obj_name + '.csv'),
+ delimiter=',', skip_header=0, names=True)
+
+ for body_param_name in self.body_sets.params_init.keys():
+ body_params[obj_name][body_param_name] = params_csv[body_param_name]
+
+ for side in ['right', 'left']:
+ wings_params[obj_name][side] = {}
+ for wing_param_name in self.wings_sets.params_init[side].keys():
+ if wing_param_name in self.body_sets.params_init.keys():
+ wings_params[obj_name][side][wing_param_name] = params_csv[wing_param_name]
+ else:
+ wings_params[obj_name][side][wing_param_name] = params_csv[wing_param_name + '_' + side]
+
+ return body_params, wings_params
+
+ def estimate_body_wings_params(self, interpolate_nans=False, filter_high_freq=False, show_plot=False):
+ body_params_labels = self.body_sets.params_init.keys()
+ wing_params_labels = self.wings_sets.params_init['right'].keys()
+
+ if len(self.skeleton3d['obj_names']) == 0:
+ raise ValueError('obj_names is empty')
+
+ self.body_params, self.wings_params, self.mean_body_params, self.mean_wings_params = {}, {}, {}, {}
+ for obj_name in self.skeleton3d['obj_names']:
+ frames = self.skeleton3d[obj_name]['frames']
+
+ self.body_params[obj_name], self.wings_params[obj_name] = {}, {}
+ self.mean_body_params[obj_name], self.mean_wings_params[obj_name] = {}, {}
+ for label in body_params_labels:
+ self.body_params[obj_name][label] = np.nan * np.ones(frames.shape)
+
+ for side in ['right', 'left']:
+ self.wings_params[obj_name][side] = {}
+ for label in wing_params_labels:
+ self.wings_params[obj_name][side][label] = np.nan * np.ones(frames.shape)
+
+ for i, frame in enumerate(frames):
+ #for label in self.skeleton3d[obj_name]['label_names']:
+
+ body_skeleton3d, wings_skeleton3d = {}, {'right': {}, 'left': {}}
+ for body_label in self.body_sets.labels:
+ body_skeleton3d[body_label] = np.array([self.skeleton3d[obj_name][body_label]['x'][i],
+ self.skeleton3d[obj_name][body_label]['y'][i],
+ self.skeleton3d[obj_name][body_label]['z'][i]])
+
+ if self.wings_sets.keep_init_aspect_ratio: init_aspect_ratio = {}
+
+ for side in ['right', 'left']:
+ for wing_label in self.wings_sets.labels:
+ label = side + '_wing_' + wing_label
+ wings_skeleton3d[side][wing_label] = \
+ np.array([self.skeleton3d[obj_name][label]['x'][i],
+ self.skeleton3d[obj_name][label]['y'][i],
+ self.skeleton3d[obj_name][label]['z'][i]])
+
+ if self.wings_sets.keep_init_aspect_ratio:
+ span_v = self.wings_skeleton3d_init[side]['tip'] - self.wings_skeleton3d_init[side]['hinge']
+ chord_v = self.wings_skeleton3d_init[side]['leading_edge_q2'] - self.wings_skeleton3d_init[side]['trailing_edge_q2']
+ init_aspect_ratio[side] = np.round(length_from_vector(span_v), 9) / np.round(length_from_vector(chord_v), 9)
+ init_aspect_ratio[side] = init_aspect_ratio[side] * np.ones(np.shape(self.wings_params[obj_name][side]['span']))
+
+ # for leg_name in leg_names:
+ # for leg_label in leg_labels:
+ # label = side + '_leg_' + leg_name + '_' + leg_label
+ # body_skeleton3d[label] = np.array([self.skeleton3d[obj_name][label]['x'][i],
+ # self.skeleton3d[obj_name][label]['y'][i],
+ # self.skeleton3d[obj_name][label]['z'][i]])
+
+ body_params = body.estimate_params_from_skeleton3d(body_skeleton3d, self.body_sets.params_init)
+ for label in body_params_labels:
+ self.body_params[obj_name][label][i] = body_params[label]
+
+ for side in ['right', 'left']:
+ wing_param = wings.estimate_params_from_skeleton3d(wings_skeleton3d[side], body_params,
+ self.wings_sets.params_init[side], side)
+
+ for label in wing_params_labels:
+ self.wings_params[obj_name][side][label][i] = wing_param[label]
+
+ if self.wings_sets.keep_init_aspect_ratio:
+ for side in ['right', 'left']:
+ self.wings_params[obj_name][side]['aspect_ratio'] = init_aspect_ratio[side]
+ self.wings_params[obj_name][side]['chord'] = self.wings_params[obj_name][side]['span']/init_aspect_ratio[side]
+
+ # if show_plot:
+ # # Plot body and wings params that will be kept cst
+ # fig1, axs1 = plt.subplots(len(self.body_sets.labels_to_keep_cst), 1)
+ # for i, label in enumerate(self.body_sets.labels_to_keep_cst):
+ # y = self.body_params[obj_name][label]
+ # axs1[i].plot(frames, y, '.', label='estim')
+ # axs1[i].plot(frames, np.nanmean(y) * np.ones(frames.shape), '-', label='mean')
+ # axs1[i].plot(frames, np.nanmedian(y) * np.ones(frames.shape), '-', label='median')
+ # axs1[i].plot(frames, stats.mode(y)[0] * np.ones(frames.shape), '-', label='mode')
+ #
+ # axs1[i].set_ylabel(label)
+ # axs1[i].set_xlabel('frames')
+ # fig1.suptitle('Body parameters in function of time')
+ # fig1.legend()
+ #
+ # fig2, axs2 = plt.subplots(len(self.wings_sets.labels_to_keep_cst), 1)
+ # for i, label in enumerate(self.wings_sets.labels_to_keep_cst):
+ # for side in ['right', 'left']:
+ # y = self.wings_params[obj_name][side][label]
+ # axs2[i].plot(frames, y, '.', label='estim_' + side)
+ # axs2[i].plot(frames, np.nanmean(y) * np.ones(frames.shape), '-', label='mean_' + side)
+ # axs2[i].plot(frames, np.nanmedian(y) * np.ones(frames.shape), '-', label='median_' + side)
+ # axs2[i].plot(frames, stats.mode(y)[0] * np.ones(frames.shape), '-', label='mode_' + side)
+ # axs2[i].set_ylabel(label)
+ # axs2[i].set_xlabel('frames')
+ # fig2.suptitle('Wings parameters in function of time')
+ # fig2.legend()
+ #
+ # plt.show()
+
+ # Compute median value of body and wings length (should be constant for one animal)
+ for label in self.body_sets.labels_to_keep_cst:
+ self.body_params[obj_name][label] = \
+ np.nanmedian(self.body_params[obj_name][label]) * np.ones(frames.shape)
+ # print('>>> mean {0}: {1}'.format(label, self.body_params[obj_name][label][0]))
+
+ self.body_params[obj_name] = filter_interp_params(self.cutoff_frequency, self.frame_rate,
+ self.body_params[obj_name], body_params_labels,
+ interpolate_nans, filter_high_freq, show_plot)
+
+ for label in self.wings_sets.labels_to_keep_cst:
+ mean_over_sides = np.nanmedian([np.nanmedian(self.wings_params[obj_name][side][label]) for side in ['right', 'left']])
+ for side in ['right', 'left']:
+ self.wings_params[obj_name][side][label] = mean_over_sides * np.ones(frames.shape)
+ # print('>>> mean {0}: {1} (right and left)'.format(label, self.wings_params[obj_name]['right'][label][0]))
+
+ for side in ['right', 'left']:
+ self.wings_params[obj_name][side] = \
+ filter_interp_params(self.cutoff_frequency, self.frame_rate, self.wings_params[obj_name][side],
+ ['x_hinge', 'y_hinge', 'z_hinge'], False, filter_high_freq, show_plot)
+ self.wings_params[obj_name][side] = \
+ filter_interp_params(self.cutoff_frequency, self.frame_rate, self.wings_params[obj_name][side],
+ wing_params_labels, interpolate_nans, False, show_plot)
+
+ # Copy all body_params in wings_params
+ for side in ['right', 'left']:
+ for label in body_params_labels:
+ self.wings_params[obj_name][side][label] = self.body_params[obj_name][label].copy()
+
+ return self.body_params, self.wings_params
+
+ def estimate_wing_hinges(self, interpolate_nans=False, filter_high_freq=False, show_plot=False):
+ wing_params_labels = self.wings_sets.params_init['right'].keys()
+
+ dpt = 3
+ x, y, z = symbols('x y z')
+
+ if len(self.skeleton3d['obj_names']) == 0:
+ raise ValueError('obj_names is empty')
+
+ for obj_name in self.skeleton3d['obj_names']:
+ frames = self.skeleton3d[obj_name]['frames']
+
+ if show_plot:
+ init_hinges = {}
+ for side in ['right', 'left']:
+ init_hinges[side] = {'x': self.wings_params[obj_name][side]['x_hinge'].copy(),
+ 'y': self.wings_params[obj_name][side]['y_hinge'].copy(),
+ 'z': self.wings_params[obj_name][side]['z_hinge'].copy()}
+
+ for side in ['right', 'left']:
+ for label in wing_params_labels:
+ self.wings_params[obj_name][side][label] = np.nan * np.ones(frames.shape)
+
+ for side in ['right', 'left']:
+ a = np.ones(np.shape(frames)) * np.nan
+ b, c, d = a.copy(), a.copy(), a.copy()
+ for i, frame in enumerate(frames):
+ x_coords = np.ones((len(self.wings_sets.labels), 1)) * np.nan
+ y_coords, z_coords = x_coords.copy(), x_coords.copy()
+ for j, wing_label in enumerate(self.wings_sets.labels):
+ label = side + '_wing_' + wing_label
+
+ x_coords[j] = self.skeleton3d[obj_name][label]['x'][i]
+ y_coords[j] = self.skeleton3d[obj_name][label]['y'][i]
+ z_coords[j] = self.skeleton3d[obj_name][label]['z'][i]
+
+ if len(x_coords) < 3: continue
+ params = minimize_perp_distance(x_coords, y_coords, z_coords)
+ a[i], b[i], c[i], d[i] = params
+
+ if i >= 2*dpt:
+ eq = [a[i - 2*dpt] * x + b[i - 2*dpt] * y + c[i - 2*dpt] * z + d[i - 2*dpt],
+ a[i - 1*dpt] * x + b[i - 1*dpt] * y + c[i - 1*dpt] * z + d[i - 1*dpt],
+ a[i] * x + b[i] * y + c[i] * z + d[i]]
+ sol = solve(eq, dict=True)
+
+ if len(sol) > 0 and x in sol[0].keys() and y in sol[0].keys() and z in sol[0].keys():
+ self.wings_params[obj_name][side]['x_hinge'][i] = sol[0][x]
+ self.wings_params[obj_name][side]['y_hinge'][i] = sol[0][y]
+ self.wings_params[obj_name][side]['z_hinge'][i] = sol[0][z]
+
+ for side in ['right', 'left']:
+ self.wings_params[obj_name][side] = \
+ filter_interp_params(self.cutoff_frequency, self.frame_rate, self.wings_params[obj_name][side],
+ ['x_hinge', 'y_hinge', 'z_hinge'], interpolate_nans, filter_high_freq,
+ show_plot)
+
+ if show_plot:
+ fig, axs = plt.subplots(1, 3)
+ for i, coord in enumerate(['x', 'y', 'z']):
+ axs[i].clear()
+ for side in ['right', 'left']:
+ axs[i].scatter(frames, init_hinges[side][coord], marker='o', label='init_' + side)
+ axs[i].scatter(frames, self.wings_params[obj_name][side][coord + '_hinge'], marker='+', label='est_' + side)
+ axs[i].set_ylabel(coord + ' pos (m)')
+ axs[i].set_xlabel('frames')
+ plt.legend()
+ plt.show()
+
+ return self.wings_params
+
+ def optim_fit_skeleton(self, animal_name, body_param_names, wing_param_names, res_method, opt_method, rec_paths,
+ save_paths, dlt_path, show_plot=False):
+ dlt_coefs = calib.read_dlt_coefs(dlt_path)
+
+ main_camn = 1
+ save_path = save_paths[main_camn]
+
+ body_params_init, wings_params_init = self.estimate_body_wings_params(interpolate_nans=True,
+ filter_high_freq=True)
+ #wings_params_init = self.estimate_wing_hinges(interpolate_nans=True, filter_high_freq=True)
+ save_skeleton_params_in_csv(self.skeleton3d, self.wings_sets.params_init, body_params_init, wings_params_init, save_path)
+
+ start = time.time()
+ body_params, nb_visible_pts_body, rmse_body, nb_iterations_body = \
+ self.optim_fit_all_hybrid_params(animal_name, body_params_init, wings_params_init, body_param_names,
+ res_method, opt_method, dlt_coefs, rec_paths, interpolate_nans=False,
+ filter_high_freq=True)
+ print('>> optim hybrid done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
+
+ # body_params, nb_visible_pts_body, rmse_body, nb_iterations_body = \
+ # self.optim_fit_all_body_params(animal_name, body_params_init, body_param_names, res_method, opt_method, dlt_coefs, rec_paths,
+ # interpolate_nans=False, filter_high_freq=True)
+ # #body_params = body_params_init
+ # print('>> optim body done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
+
+ # self.unscramble_sides3d_dlc(body_params)
+ # self.save2d_dlc(save_paths)
+ # self.save3d_dlc(save_paths)
+
+ start = time.time()
+ wings_params, nb_visible_pts_wings, rmse_wings, nb_iterations_wings = \
+ self.optim_fit_all_wings_params(animal_name, body_params, wings_params_init, wing_param_names, res_method,
+ opt_method, dlt_coefs, rec_paths, interpolate_nans=False,
+ filter_high_freq=True)
+ print('>> optim wings done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
+
+ #body_params, wings_params = self.correct_body_wings_params(body_params, wings_params, angles_to_correct, rec_paths, dlt_coefs)
+
+ if show_plot:
+ for obj_name in body_params.keys():
+ frames = list(range(0, len(body_params[obj_name][body_param_names[0]])))
+
+ # Plot body params
+ fig1, axs1 = plt.subplots(len(body_param_names), 1)
+ for i, label in enumerate(body_param_names):
+ axs1[i].plot(frames, body_params_init[obj_name][label], '-', label='init')
+ axs1[i].plot(frames, body_params[obj_name][label], '-', label='optim')
+ axs1[i].set_ylabel(label)
+ axs1[i].set_xlabel('frames')
+ fig1.suptitle('Body parameters in function of time')
+ fig1.legend()
+
+ # Plot wings params
+ fig2, axs2 = plt.subplots(len(wing_param_names), 1)
+ for i, label in enumerate(wing_param_names):
+ for side in wings_params[obj_name].keys():
+ axs2[i].plot(frames, wings_params_init[obj_name][side][label], '-', label='init_' + side)
+ axs2[i].plot(frames, wings_params[obj_name][side][label], '-', label='optim_' + side)
+ axs2[i].set_ylabel(label)
+ axs2[i].set_xlabel('frames')
+ fig2.suptitle('Wings parameters in function of time')
+ fig2.legend()
+
+ plt.show()
+
+ save_skeleton_params_in_csv(self.skeleton3d, self.wings_sets.params_init, body_params, wings_params, save_path)
+
+ save_obj_metric_in_csv(self.skeleton3d, self.wings_sets.params_init, nb_visible_pts_body, 'nb_visible_pts-body', save_path)
+ save_obj_metric_in_csv(self.skeleton3d, self.wings_sets.params_init, nb_visible_pts_wings, 'nb_visible_pts-wings', save_path)
+ save_obj_metric_in_csv(self.skeleton3d, self.wings_sets.params_init, nb_iterations_body, 'nb_iterations-body', save_path)
+ save_obj_metric_in_csv(self.skeleton3d, self.wings_sets.params_init, nb_iterations_wings, 'nb_iterations-wings', save_path)
+ save_obj_metric_in_csv(self.skeleton3d, self.wings_sets.params_init, rmse_body, 'rmse-body', save_path)
+ save_obj_metric_in_csv(self.skeleton3d, self.wings_sets.params_init, rmse_wings, 'rmse-wings', save_path)
+
+ def optim_fit_all_body_params(self, animal_name, body_params_init, param_names, res_method, opt_method, dlt_coefs,
+ rec_paths, interpolate_nans=False, filter_high_freq=False, show_plot=False):
+
+ res_method = 'body_' + res_method
+
+ nb_cam = len(rec_paths.keys())
+
+ body_param_ests, nb_visible_pts, rmse, nb_iterations = {}, {}, {}, {}
+ obj_names = self.skeleton2d['obj_names'] if '2d' in res_method else self.skeleton3d['obj_names']
+ for obj_name in obj_names:
+ frames = self.skeleton2d[obj_name]['frames'] if '2d' in res_method else self.skeleton3d[obj_name]['frames']
+ # frames = [frame for i, frame in enumerate(frames) if i % 100 == 0.0]
+
+ shape_frame = np.array(frames).shape
+ nb_iterations[obj_name] = np.zeros(shape_frame)
+ rmse[obj_name], nb_visible_pts[obj_name] = np.ones(shape_frame) * np.nan, np.ones(shape_frame) * np.nan
+ if obj_name in body_params_init.keys():
+ body_param_ests[obj_name] = copy.deepcopy(body_params_init[obj_name])
+ else:
+ body_param_ests[obj_name] = {}
+ for label in self.body_sets.params_init.keys():
+ body_param_ests[obj_name][label] = np.ones(shape_frame) * self.body_sets.params_init[label]
+
+ if show_plot and not self.multiprocessing:
+ fig, axs = plt.subplots(1, nb_cam)
+ all_images = {}
+ for camn in range (1, nb_cam +1):
+ all_files = sorted(os.listdir(rec_paths[camn]), key=lambda s: s[s.rfind('.') - nb_leading_zeros:s.rfind('.')])
+ all_images[camn] = [s for s in all_files if '.{0}'.format(image_format) in s]
+
+ if self.multiprocessing: args = [(frame,) for frame in frames]
+
+ #for i, frame in enumerate(tqdm.tqdm(frames)): # to print progress bar
+ for i, frame in enumerate(frames):
+ if obj_name in body_params_init.keys():
+ body_params_init_frame = {}
+ for label in self.body_sets.params_init.keys():
+ body_params_init_frame[label] = body_params_init[obj_name][label][i]
+ else:
+ body_params_init_frame = self.body_sets.params_init
+
+ nb_points = len(self.body_skeleton3d_init.keys())
+ nb_visible_pts[obj_name][i] = nb_points
+ if '3d' in res_method:
+ body_skeleton3d = {}
+ for body_label in self.body_skeleton3d_init.keys():
+ body_skeleton3d[body_label] = \
+ np.array([self.skeleton3d[obj_name][body_label]['x'][i],
+ self.skeleton3d[obj_name][body_label]['y'][i],
+ self.skeleton3d[obj_name][body_label]['z'][i]])
+
+ if np.isnan(body_skeleton3d[body_label]).any(): nb_visible_pts[obj_name][i] -= 1
+
+ elif '2d' in res_method:
+ body_skeleton2d = {}
+ for body_label in self.body_skeleton3d_init.keys():
+ body_skeleton2d[body_label] = {1: np.array([])}
+ for camn in range(1, nb_cam + 1):
+
+ if self.skeleton2d[obj_name][body_label][camn]['likelihood'][i] > self.body_sets.threshold_likelihood:
+ body_skeleton2d[body_label][camn] = \
+ np.array([self.skeleton2d[obj_name][body_label][camn]['x'][i],
+ self.skeleton2d[obj_name][body_label][camn]['y'][i],
+ self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
+ else:
+ body_skeleton2d[body_label][camn] = \
+ np.array([np.nan, np.nan, self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
+
+ nb_visible_pts[obj_name][i] -= 1 / nb_cam
+
+ return_nan = nb_visible_pts[obj_name][i] <= self.body_sets.threshold_nb_pts
+
+ if not self.multiprocessing:
+ if '3d' in res_method:
+ body_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
+ optim_fit_body_params(animal_name, self.body_skeleton3d_init, body_skeleton3d, body_params_init_frame,
+ param_names, res_method, opt_method, self.body_sets.bounds_init, return_nan)
+
+ elif '2d' in res_method:
+ body_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
+ optim_fit_body_params(animal_name, self.body_skeleton3d_init, body_skeleton2d, body_params_init_frame,
+ param_names, res_method, opt_method, self.body_sets.bounds_init, return_nan, dlt_coefs=dlt_coefs)
+
+ for label in body_param_ests_frame.keys():
+ body_param_ests[obj_name][label][i] = body_param_ests_frame[label]
+
+ if show_plot and (frame % 20 == 0.0 or frame == 1):
+ body_skeleton3d_init = copy.deepcopy(self.body_skeleton3d_init)
+ body_skeleton3d_init = body.build_skeleton3d(body_skeleton3d_init, body_params_init_frame)
+ body_skeleton3d_init = body.rotate_skeleton3d(body_skeleton3d_init, body_params_init_frame)
+ body_skeleton3d_init = body.translate_skeleton3d(body_skeleton3d_init, body_params_init_frame)
+ body_skeleton2d_init = reproject_skeleton3d_to2d(body_skeleton3d_init, dlt_coefs)
+
+ body_skeleton3d_new = copy.deepcopy(self.body_skeleton3d_init)
+ body_skeleton3d_new = body.build_skeleton3d(body_skeleton3d_new, body_param_ests_frame)
+ body_skeleton3d_new = body.rotate_skeleton3d(body_skeleton3d_new, body_param_ests_frame)
+ body_skeleton3d_new = body.translate_skeleton3d(body_skeleton3d_new, body_param_ests_frame)
+ body_skeleton2d_new = reproject_skeleton3d_to2d(body_skeleton3d_new, dlt_coefs)
+
+ # Compute reprojection of 3d reconstruction from dlc 2d points
+ body_skeleton3d_repro2d = {}
+ for label in self.body_skeleton3d_init.keys():
+ body_skeleton3d_repro2d[label] = {1: []}
+ for j, camn in enumerate(range(1, nb_cam + 1)):
+ uv = calib.find2d(1, dlt_coefs[j], np.array([[self.skeleton3d[obj_name][label]['x'][i],
+ self.skeleton3d[obj_name][label]['y'][i], self.skeleton3d[obj_name][label]['z'][i]]]))
+ body_skeleton3d_repro2d[label][camn] = np.array([uv[0][0], uv[0][1], 1.0])
+
+ if '3d' in res_method:
+ body_skeleton2d = reproject_skeleton3d_to2d(body_skeleton3d, dlt_coefs)
+
+ # Plot 2d repojection of body skeleton
+ for j, camn in enumerate(range(1, nb_cam +1)):
+ center_of_mass_xy = \
+ (body_skeleton2d_new['torso_abdomen_joint'][camn] * body_param_ests_frame['ratio_com_torso'] \
+ + body_skeleton2d_new['head_torso_joint'][camn] * (1 - body_param_ests_frame['ratio_com_torso']))
+
+ if np.isnan(center_of_mass_xy).any(): continue
+
+ axs[j].clear()
+ axs[j].imshow(Image.open(os.path.join(rec_paths[camn], all_images[camn][i])))
+ for label in body_skeleton2d_new.keys():
+ if body_skeleton2d[label][camn][2] > self.body_sets.threshold_likelihood:
+ axs[j].scatter(body_skeleton2d[label][camn][0],
+ body_skeleton2d[label][camn][1], marker='.', color='k')
+ axs[j].scatter(body_skeleton3d_repro2d[label][camn][0],
+ body_skeleton3d_repro2d[label][camn][1], marker='o', facecolors='none', edgecolors='k')
+ axs[j].scatter(body_skeleton2d_init[label][camn][0],
+ body_skeleton2d_init[label][camn][1], marker='o', facecolors='none', edgecolors='b')
+ axs[j].scatter(body_skeleton2d_new[label][camn][0],
+ body_skeleton2d_new[label][camn][1], marker='+', color='r')
+
+ axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
+ axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
+
+ plt.pause(0.05)
+ # plt.show()
+
+ # param_to_print = ['pitch_a', 'roll_a', 'yaw_a', 'ratio_body']
+ # for label in param_to_print:
+ # print('{0}: init: {1} and est: {2}'.format(label, body_params_init_frame[label], body_param_ests_frame[label]))
+
+ else:
+ if '3d' in res_method:
+ args[i] = (self.body_skeleton3d_init, body_skeleton3d, body_params_init_frame, param_names,
+ res_method, opt_method, self.body_sets.bounds_init, return_nan,)
+
+ elif '2d' in res_method:
+ args[i] = (self.body_skeleton3d_init, body_skeleton2d, body_params_init_frame, param_names,
+ res_method, opt_method, self.body_sets.bounds_init, return_nan,)
+
+ if self.multiprocessing:
+ pool_body = Pool(self.nb_process)
+ if '3d' in res_method:
+ results = pool_body.starmap(optim_fit_body_params, args)
+
+ elif '2d' in res_method:
+ results = pool_body.starmap(partial(optim_fit_body_params, dlt_coefs=dlt_coefs), args)
+
+ pool_body.close()
+ for i, frame in enumerate(frames):
+ body_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = results[i]
+ for label in body_param_ests_frame.keys():
+ body_param_ests[obj_name][label][i] = body_param_ests_frame[label]
+
+ body_param_ests[obj_name] = filter_interp_params(self.cutoff_frequency, self.frame_rate,
+ body_param_ests[obj_name], param_names, interpolate_nans,
+ filter_high_freq, show_plot)
+
+ return body_param_ests, nb_visible_pts, rmse, nb_iterations
+
+ def optim_fit_all_wings_params(self, animal_name, body_params_init, wings_params_init, param_names, res_method, opt_method,
+ dlt_coefs, rec_paths, interpolate_nans=False, filter_high_freq=False, show_plot=False):
+
+ res_method = 'wing_' + res_method
+
+ nb_cam = len(rec_paths.keys())
+
+ obj_names = self.skeleton2d['obj_names'] if '2d' in res_method else self.skeleton3d['obj_names']
+ for obj_name in obj_names:
+ if obj_name in wings_params_init.keys():
+ for side in self.wings_sets.params_init.keys():
+ for label in body_params_init[obj_name].keys():
+ wings_params_init[obj_name][side][label] = body_params_init[obj_name][label]
+
+ wings_params_ests, nb_visible_pts, rmse, nb_iterations = copy.deepcopy(wings_params_init), {}, {}, {}
+ for obj_name in obj_names:
+ frames = self.skeleton2d[obj_name]['frames'] if '2d' in res_method else self.skeleton3d[obj_name]['frames']
+ # frames = [frame for i, frame in enumerate(frames) if i % 100 == 0.0]
+
+ if show_plot and not self.multiprocessing:
+ fig, axs = plt.subplots(1, nb_cam)
+ all_images = {}
+ for camn in range(1, nb_cam + 1):
+ all_files = sorted(os.listdir(rec_paths[camn]), key=lambda s: s[s.rfind('.') - nb_leading_zeros:s.rfind('.')])
+ all_images[camn] = [s for s in all_files if '.{0}'.format(image_format) in s]
+
+ if self.multiprocessing: args = [(frame,) for frame in frames]
+
+ shape_frame = np.array(frames).shape
+ rmse[obj_name] = {'right': np.ones(shape_frame) * np.nan, 'left': np.ones(shape_frame) * np.nan}
+ nb_iterations[obj_name] = {'right': np.zeros(shape_frame), 'left': np.zeros(shape_frame)}
+ nb_visible_pts[obj_name] = {'right': np.ones(shape_frame) * np.nan, 'left': np.ones(shape_frame) * np.nan}
+ #for i, frame in enumerate(tqdm.tqdm(frames)): # to print progress bar
+ for i, frame in enumerate(frames):
+ if obj_name in wings_params_init.keys():
+ wings_params_init_frame = {}
+ for side in self.wings_sets.params_init.keys():
+ wings_params_init_frame[side] = {}
+ for label in self.wings_sets.params_init[side].keys():
+ wings_params_init_frame[side][label] = wings_params_init[obj_name][side][label][i]
+ else:
+ wings_params_init_frame = self.wings_sets.params_init
+
+ nb_points = len(self.body_skeleton3d_init.keys())
+ for side in self.wings_skeleton3d_init.keys():
+ nb_visible_pts[obj_name][side][i] = nb_points
+
+ if '3d' in res_method:
+ wings_skeleton3d = {'right': {}, 'left': {}}
+ for side in self.wings_skeleton3d_init.keys():
+ for wing_label in self.wings_skeleton3d_init[side].keys():
+ label = side + '_wing_' + wing_label
+ wings_skeleton3d[side][wing_label] = \
+ np.array([self.skeleton3d[obj_name][label]['x'][i],
+ self.skeleton3d[obj_name][label]['y'][i],
+ self.skeleton3d[obj_name][label]['z'][i]])
+
+ if np.isnan(wings_skeleton3d[side][wing_label]).any():
+ nb_visible_pts[obj_name][side][i] -= 1
+
+ elif '2d' in res_method:
+ wings_skeleton2d = {'right': {}, 'left': {}}
+ for side in self.wings_skeleton3d_init.keys():
+ for wing_label in self.wings_skeleton3d_init[side].keys():
+
+ wings_skeleton2d[side][wing_label] = {1: []}
+ for camn in range(1, nb_cam + 1):
+ label = side + '_wing_' + wing_label
+
+ if self.skeleton2d[obj_name][label][camn]['likelihood'][i] > self.wings_sets.threshold_likelihood:
+ wings_skeleton2d[side][wing_label][camn] = \
+ np.array([self.skeleton2d[obj_name][label][camn]['x'][i],
+ self.skeleton2d[obj_name][label][camn]['y'][i],
+ self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
+ else:
+ wings_skeleton2d[side][wing_label][camn] = \
+ np.array([np.nan, np.nan, self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
+
+ nb_visible_pts[obj_name][side][i] -= 1 / nb_cam
+
+ body_params_frame = {}
+ for label in body_params_init[obj_name].keys():
+ body_params_frame[label] = body_params_init[obj_name][label][i]
+
+ body_skeleton3d = copy.deepcopy(self.body_skeleton3d_init)
+ body_skeleton3d = body.build_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton3d = body.rotate_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton3d = body.translate_skeleton3d(body_skeleton3d, body_params_frame)
+
+ return_nan = [False, False]
+ for j, side in enumerate(self.wings_skeleton3d_init.keys()):
+ [wings_params_init_frame[side]['x_hinge'], wings_params_init_frame[side]['y_hinge'],
+ wings_params_init_frame[side]['z_hinge']] = body_skeleton3d['{0}_wing_hinge'.format(side)]
+
+ return_nan[j] = nb_visible_pts[obj_name][side][i] <= self.wings_sets.threshold_nb_pts
+ if return_nan[j]:
+ for label in param_names:
+ wings_params_ests[obj_name][side][label][i] = np.nan
+ continue
+
+ for label in ['yaw_a', 'pitch_a', 'roll_a', 'x_com', 'y_com', 'z_com']:
+ wings_params_init_frame[side][label] = body_params_frame[label]
+
+ if not self.multiprocessing:
+ if 'geo' in res_method:
+ if '3d' in res_method:
+ wings_params_est_frame, rmse_frame, nb_iterations_frame = \
+ optim_fit_limbs_params(animal_name, self.wings_geometry3d_init, wings_skeleton3d, wings_params_init_frame,
+ param_names, res_method, opt_method, self.wings_sets.bounds_init, return_nan)
+
+ elif '2d' in res_method:
+ wings_params_est_frame, rmse_frame, nb_iterations_frame = \
+ optim_fit_limbs_params(animal_name, self.wings_geometry3d_init, wings_skeleton2d, wings_params_init_frame,
+ param_names, res_method, opt_method, self.wings_sets.bounds_init, return_nan,
+ dlt_coefs=dlt_coefs)
+ else:
+ if '3d' in res_method:
+ wings_params_est_frame, rmse_frame, nb_iterations_frame = \
+ optim_fit_limbs_params(animal_name, self.wings_skeleton3d_init, wings_skeleton3d, wings_params_init_frame,
+ param_names, res_method, opt_method, self.wings_sets.bounds_init, return_nan)
+
+ elif '2d' in res_method:
+ wings_params_est_frame, rmse_frame, nb_iterations_frame = \
+ optim_fit_limbs_params(animal_name, self.wings_skeleton3d_init, wings_skeleton2d, wings_params_init_frame,
+ param_names, res_method, opt_method, self.wings_sets.bounds_init, return_nan,
+ dlt_coefs=dlt_coefs)
+
+ for side in self.wings_skeleton3d_init.keys():
+ rmse[obj_name][side][i] = rmse_frame[side]
+ nb_iterations[obj_name][side][i] = nb_iterations_frame[side]
+ for label in wings_params_est_frame[side].keys():
+ wings_params_ests[obj_name][side][label][i] = wings_params_est_frame[side][label]
+
+ if show_plot and frame % 20 == 0.0:
+ body_skeleton2d_new = reproject_skeleton3d_to2d(body_skeleton3d, dlt_coefs)
+
+ wings_skeleton3d_new = copy.deepcopy(self.wings_skeleton3d_init)
+ wings_skeleton3d_init = copy.deepcopy(self.wings_skeleton3d_init)
+ wings_skeleton2d_new, wings_skeleton2d_init = {'right': {}, 'left': {}}, {'right': {}, 'left': {}}
+ wings_skeleton3d_repro2d = {'right': {}, 'left': {}}
+ if '3d' in res_method: wings_skeleton2d = copy.deepcopy(wings_skeleton2d_new)
+ for side in self.wings_skeleton3d_init.keys():
+ wings_skeleton3d_init[side] = wings.build_skeleton3d(wings_skeleton3d_init[side], wings_params_init_frame[side])
+ wings_skeleton3d_init[side] = wings.rotate_and_translate_skeleton3d(wings_skeleton3d_init[side], wings_params_init_frame[side], side)
+ wings_skeleton2d_init[side] = reproject_skeleton3d_to2d(wings_skeleton3d_init[side], dlt_coefs)
+
+ wings_skeleton3d_new[side] = wings.build_skeleton3d(wings_skeleton3d_new[side], wings_params_est_frame[side])
+ wings_skeleton3d_new[side] = wings.rotate_and_translate_skeleton3d(wings_skeleton3d_new[side], wings_params_est_frame[side], side)
+ wings_skeleton2d_new[side] = reproject_skeleton3d_to2d(wings_skeleton3d_new[side], dlt_coefs)
+
+ # Compute reprojection of 3d reconstruction from dlc 2d points
+ for label in self.wings_skeleton3d_init[side].keys():
+ wings_skeleton3d_repro2d[side][label] = {1: []}
+ for j, camn in enumerate(range(1, nb_cam + 1)):
+ label2 = side + '_wing_' + label
+ uv = calib.find2d(1, dlt_coefs[j], np.array([[self.skeleton3d[obj_name][label2]['x'][i],
+ self.skeleton3d[obj_name][label2]['y'][i], self.skeleton3d[obj_name][label2]['z'][i]]]))
+ wings_skeleton3d_repro2d[side][label][camn] = np.array([uv[0][0], uv[0][1], 1.0])
+
+ if '3d' in res_method:
+ wings_skeleton2d[side] = reproject_skeleton3d_to2d(wings_skeleton3d[side], dlt_coefs)
+
+ for j, camn in enumerate(range(1, nb_cam + 1)):
+ center_of_mass_xy = \
+ (body_skeleton2d_new['torso_abdomen_joint'][camn] * body_params_frame['ratio_com_torso'] \
+ + body_skeleton2d_new['head_torso_joint'][camn] * (1 - body_params_frame['ratio_com_torso']))
+
+ if np.isnan(center_of_mass_xy).any(): continue
+
+ axs[j].clear()
+ axs[j].imshow(Image.open(os.path.join(rec_paths[camn], all_images[camn][i])))
+ for side in self.wings_skeleton3d_init.keys():
+ for label in wings_skeleton2d_new[side].keys():
+ if wings_skeleton2d[side][label][camn][2] > self.wings_sets.threshold_likelihood:
+ axs[j].scatter(wings_skeleton2d[side][label][camn][0],
+ wings_skeleton2d[side][label][camn][1], marker='.', color='k')
+ axs[j].scatter(wings_skeleton3d_repro2d[side][label][camn][0],
+ wings_skeleton3d_repro2d[side][label][camn][1], marker='o', facecolors='none', edgecolors='k')
+ axs[j].scatter(wings_skeleton2d_init[side][label][camn][0],
+ wings_skeleton2d_init[side][label][camn][1], marker='o', facecolors='none', edgecolors='b')
+ axs[j].scatter(wings_skeleton2d_new[side][label][camn][0],
+ wings_skeleton2d_new[side][label][camn][1], marker='+', color='r')
+
+ axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
+ axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
+
+ plt.pause(0.05)
+ #plt.show()
+
+ # param_to_print = ['stroke_a', 'deviation_a', 'rotation_a', 'ratio_wing']
+ # for label in param_to_print:
+ # print('{0}: init: {1} and est: {2}'.format(label, wings_params_init_frame[side][label],
+ # wings_params_est_frame[side][label]))
+
+ else:
+ if 'geo' in res_method:
+ if '3d' in res_method:
+ args[i] = (self.wings_geometry3d_init, wings_skeleton3d, wings_params_init_frame, param_names,
+ res_method, opt_method, self.wings_sets.bounds_init, return_nan)
+
+ elif '2d' in res_method:
+ args[i] = (self.wings_geometry3d_init, wings_skeleton2d, wings_params_init_frame, param_names,
+ res_method, opt_method, self.wings_sets.bounds_init, return_nan)
+ else:
+ if '3d' in res_method:
+ args[i] = (self.wings_skeleton3d_init, wings_skeleton3d, wings_params_init_frame, param_names,
+ res_method, opt_method, self.wings_sets.bounds_init, return_nan)
+
+ elif '2d' in res_method:
+ args[i] = (self.wings_skeleton3d_init, wings_skeleton2d, wings_params_init_frame, param_names,
+ res_method, opt_method, self.wings_sets.bounds_init, return_nan)
+
+ if self.multiprocessing:
+ pool_wings = Pool(self.nb_process)
+ if '3d' in res_method:
+ results = pool_wings.starmap(optim_fit_limbs_params, args)
+
+ elif '2d' in res_method:
+ results = pool_wings.starmap(partial(optim_fit_limbs_params, dlt_coefs=dlt_coefs), args)
+
+ pool_wings.close()
+ for i, frame in enumerate(frames):
+ wings_params_est_frame, rmse_frame, nb_iterations_frame = results[i]
+ for side in self.wings_skeleton3d_init.keys():
+ rmse[obj_name][side][i] = rmse_frame[side]
+ nb_iterations[obj_name][side][i] = nb_iterations_frame[side]
+ for label in wings_params_est_frame[side].keys():
+ wings_params_ests[obj_name][side][label][i] = wings_params_est_frame[side][label]
+
+ for side in self.wings_skeleton3d_init.keys():
+ wings_params_ests[obj_name][side] = \
+ filter_interp_params(self.cutoff_frequency, self.frame_rate, wings_params_ests[obj_name][side],
+ ['x_hinge', 'y_hinge', 'z_hinge'], False, filter_high_freq, show_plot)
+ wings_params_ests[obj_name][side] = \
+ filter_interp_params(self.cutoff_frequency, self.frame_rate, wings_params_ests[obj_name][side],
+ param_names, interpolate_nans, False, show_plot)
+
+ return wings_params_ests, nb_visible_pts, rmse, nb_iterations
+
+ def optim_fit_all_hybrid_params(self, animal_name, body_params_init, wings_params_init, param_names, res_method,
+ opt_method, dlt_coefs, rec_paths, interpolate_nans=False, filter_high_freq=False,
+ first_optim_wings=False, show_plot=False):
+
+ nb_cam = len(rec_paths.keys())
+
+ hybrid_params_init = copy.deepcopy(body_params_init)
+ if '2d' in res_method: self.hybrid_skeleton2d = copy.deepcopy(self.skeleton2d)
+ else: self.hybrid_skeleton3d = copy.deepcopy(self.skeleton3d)
+
+ if first_optim_wings:
+ start = time.time()
+ # First optim on wing to estimate wing tips and hinges positions + deviation_a
+ wing_param_names = ['stroke_a', 'deviation_a', 'rotation_a', 'x_hinge', 'y_hinge', 'z_hinge']
+ wings_params, _, _ = \
+ self.optim_fit_all_wings_params(body_params_init, wings_params_init, wing_param_names, res_method,
+ opt_method, dlt_coefs, rec_paths, interpolate_nans=False, filter_high_freq=True, show_plot=show_plot)
+
+ obj_names = self.skeleton2d['obj_names'] if '2d' in res_method else self.skeleton3d['obj_names']
+ for obj_name in obj_names:
+
+ frames = self.skeleton2d[obj_name]['frames'] if '2d' in res_method else self.skeleton3d[obj_name]['frames']
+ for i, frame in enumerate(frames):
+
+ wings_params_frame = {'right': {}, 'left': {}}
+ for side in self.wings_skeleton3d_init.keys():
+ for label in wings_params[obj_name][side].keys():
+ wings_params_frame[side][label] = wings_params[obj_name][side][label][i]
+
+ wings_skeleton3d_new = copy.deepcopy(self.wings_skeleton3d_init)
+ wings_skeleton2d_new = {'right': {}, 'left': {}}
+ for side in self.wings_skeleton3d_init.keys():
+ wings_skeleton3d_new[side] = wings.build_skeleton3d(wings_skeleton3d_new[side], wings_params_frame[side])
+ wings_skeleton3d_new[side] = \
+ wings.rotate_and_translate_skeleton3d(wings_skeleton3d_new[side], wings_params_frame[side], side)
+
+ if '2d' in res_method:
+ wings_skeleton2d_new[side] = reproject_skeleton3d_to2d(wings_skeleton3d_new[side], dlt_coefs)
+
+ for label in ['tip', 'hinge']:
+ wing_label = side + '_wing_' + label
+
+ if '2d' in res_method:
+ for camn in range(1, nb_cam + 1):
+ self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'][i] = wings_skeleton2d_new[side][label][camn][0]
+ self.hybrid_skeleton2d[obj_name][wing_label][camn]['y'][i] = wings_skeleton2d_new[side][label][camn][1]
+ self.hybrid_skeleton2d[obj_name][wing_label][camn]['likelihood'][i] = 1.0
+
+ else:
+ self.hybrid_skeleton3d[obj_name][wing_label]['x'][i] = wings_skeleton3d_new[side][label][0]
+ self.hybrid_skeleton3d[obj_name][wing_label]['y'][i] = wings_skeleton3d_new[side][label][1]
+ self.hybrid_skeleton3d[obj_name][wing_label]['z'][i] = wings_skeleton3d_new[side][label][2]
+
+ print('>> first optim wing done! (time elapsed: {0:.4f} s)'.format(time.time() - start))
+
+ res_method = 'hybrid_' + res_method
+
+ hybrid_param_ests, nb_visible_pts, rmse, nb_iterations = {}, {}, {}, {}
+ obj_names = self.hybrid_skeleton2d['obj_names'] if '2d' in res_method else self.hybrid_skeleton3d['obj_names']
+ for obj_name in obj_names:
+ frames = self.hybrid_skeleton2d[obj_name]['frames'] if '2d' in res_method else self.hybrid_skeleton3d[obj_name]['frames']
+ # frames = [frame for i, frame in enumerate(frames) if i % 100 == 0.0]
+
+ # Low pass filter the position of wing tips and hinges + deviation_a and span
+ for param_name in ['deviation_a', 'span']:
+ hybrid_params_init[obj_name]['wbaverage_' + param_name] = \
+ np.mean([np.transpose(filter_interp(wings_params_init[obj_name]['right'][param_name],
+ self.frame_rate, self.cutoff_frequency)),
+ np.transpose(filter_interp(wings_params_init[obj_name]['left'][param_name],
+ self.frame_rate, self.cutoff_frequency))], axis=0)
+
+ for side in self.wings_sets.params_init.keys():
+ for label in ['tip', 'hinge']:
+ wing_label = side +'_wing_' + label
+ if '2d' in res_method:
+ for camn in range(1, nb_cam + 1):
+ self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'] = \
+ filter_interp(self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'],
+ self.frame_rate, self.cutoff_frequency)
+ self.hybrid_skeleton2d[obj_name][wing_label][camn]['y'] = \
+ filter_interp(self.hybrid_skeleton2d[obj_name][wing_label][camn]['y'],
+ self.frame_rate, self.cutoff_frequency)
+ self.hybrid_skeleton2d[obj_name][wing_label][camn]['likelihood'] = \
+ np.ones(np.size(self.hybrid_skeleton2d[obj_name][wing_label][camn]['x'],
+ self.frame_rate, self.cutoff_frequency))
+
+ else:
+ self.hybrid_skeleton3d[obj_name][wing_label]['x'] = \
+ filter_interp(self.hybrid_skeleton3d[obj_name][wing_label]['x'],
+ self.frame_rate, self.cutoff_frequency)
+ self.hybrid_skeleton3d[obj_name][wing_label]['y'] = \
+ filter_interp(self.hybrid_skeleton3d[obj_name][wing_label]['y'],
+ self.frame_rate, self.cutoff_frequency)
+ self.hybrid_skeleton3d[obj_name][wing_label]['z'] = \
+ filter_interp(self.hybrid_skeleton3d[obj_name][wing_label]['z'],
+ self.frame_rate, self.cutoff_frequency)
+
+ shape_frame = np.array(frames).shape
+ nb_iterations[obj_name] = np.zeros(shape_frame)
+ nb_visible_pts[obj_name], rmse[obj_name] = np.ones(shape_frame) * np.nan, np.ones(shape_frame) * np.nan
+ if obj_name in hybrid_params_init.keys():
+ hybrid_param_ests[obj_name] = copy.deepcopy(hybrid_params_init[obj_name])
+ else:
+ hybrid_param_ests[obj_name] = {}
+ for label in self.hybrid_sets.params_init.keys():
+ hybrid_param_ests[obj_name][label] = np.ones(shape_frame) * self.hybrid_sets.params_init[label]
+
+ if show_plot and not self.multiprocessing:
+ fig, axs = plt.subplots(1, nb_cam)
+ all_images = {}
+ for camn in range (1, nb_cam +1):
+ all_files = sorted(os.listdir(rec_paths[camn]), key=lambda s: s[s.rfind('.') - nb_leading_zeros:s.rfind('.')])
+ all_images[camn] = [s for s in all_files if '.{0}'.format(image_format) in s]
+
+ if self.multiprocessing: args = [(frame,) for frame in frames]
+
+ #for i, frame in enumerate(tqdm.tqdm(frames)): # to print progress bar
+ for i, frame in enumerate(frames):
+ if obj_name in hybrid_params_init.keys():
+ hybrid_params_init_frame = {}
+ for label in self.hybrid_sets.params_init.keys():
+ hybrid_params_init_frame[label] = hybrid_params_init[obj_name][label][i]
+ else:
+ hybrid_params_init_frame = self.hybrid_sets.params_init
+
+ nb_points = len(self.hybrid_sets.skeleton3d_init.keys())
+ nb_visible_pts[obj_name][i] = nb_points
+ if '3d' in res_method:
+ hybrid_skeleton3d = {}
+ for hybrid_label in self.hybrid_sets.skeleton3d_init.keys():
+ hybrid_skeleton3d[hybrid_label] = \
+ np.array([self.hybrid_skeleton3d[obj_name][hybrid_label]['x'][i],
+ self.hybrid_skeleton3d[obj_name][hybrid_label]['y'][i],
+ self.hybrid_skeleton3d[obj_name][hybrid_label]['z'][i]])
+
+ if np.isnan(hybrid_skeleton3d[hybrid_label]).any(): nb_visible_pts[obj_name][i] -= 1
+
+ elif '2d' in res_method:
+ hybrid_skeleton2d = {}
+ for hybrid_label in self.hybrid_sets.skeleton3d_init.keys():
+ hybrid_skeleton2d[hybrid_label] = {1: np.array([])}
+ for camn in range(1, nb_cam + 1):
+
+ if self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['likelihood'][i] > self.hybrid_sets.threshold_likelihood:
+ hybrid_skeleton2d[hybrid_label][camn] = \
+ np.array([self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['x'][i],
+ self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['y'][i],
+ self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['likelihood'][i]])
+ else:
+ hybrid_skeleton2d[hybrid_label][camn] = \
+ np.array([np.nan, np.nan, self.hybrid_skeleton2d[obj_name][hybrid_label][camn]['likelihood'][i]])
+
+ nb_visible_pts[obj_name][i] -= 1 / nb_cam
+
+ return_nan = nb_visible_pts[obj_name][i] <= self.body_sets.threshold_nb_pts - 2 # to compensate for the extra wing tips
+
+ if not self.multiprocessing:
+ if '3d' in res_method:
+ hybrid_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
+ optim_fit_body_params(animal_name, self.hybrid_sets.skeleton3d_init, hybrid_skeleton3d, hybrid_params_init_frame,
+ param_names, res_method, opt_method, self.hybrid_bounds_init, return_nan)
+
+ elif '2d' in res_method:
+ hybrid_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = \
+ optim_fit_body_params(animal_name, self.hybrid_sets.skeleton3d_init, hybrid_skeleton2d, hybrid_params_init_frame,
+ param_names, res_method, opt_method, self.hybrid_bounds_init, return_nan, dlt_coefs=dlt_coefs)
+
+ for label in hybrid_param_ests_frame.keys():
+ hybrid_param_ests[obj_name][label][i] = hybrid_param_ests_frame[label]
+
+ if show_plot and (frame % 20 == 0.0 or frame == 1):
+ hybrid_skeleton3d_init = copy.deepcopy(self.hybrid_sets.skeleton3d_init)
+ hybrid_skeleton3d_init = hybrid.build_skeleton3d(hybrid_skeleton3d_init, hybrid_params_init_frame)
+ hybrid_skeleton3d_init = body.rotate_skeleton3d(hybrid_skeleton3d_init, hybrid_params_init_frame)
+ hybrid_skeleton3d_init = body.translate_skeleton3d(hybrid_skeleton3d_init, hybrid_params_init_frame)
+ hybrid_skeleton2d_init = reproject_skeleton3d_to2d(hybrid_skeleton3d_init, dlt_coefs)
+
+ hybrid_skeleton3d_new = copy.deepcopy(self.hybrid_sets.skeleton3d_init)
+ hybrid_skeleton3d_new = hybrid.build_skeleton3d(hybrid_skeleton3d_new, hybrid_param_ests_frame)
+ hybrid_skeleton3d_new = body.rotate_skeleton3d(hybrid_skeleton3d_new, hybrid_param_ests_frame)
+ hybrid_skeleton3d_new = body.translate_skeleton3d(hybrid_skeleton3d_new, hybrid_param_ests_frame)
+ hybrid_skeleton2d_new = reproject_skeleton3d_to2d(hybrid_skeleton3d_new, dlt_coefs)
+
+ # Compute reprojection of 3d reconstruction from dlc 2d points
+ hybrid_skeleton3d_repro2d = {}
+ for label in self.hybrid_sets.skeleton3d_init.keys():
+ hybrid_skeleton3d_repro2d[label] = {1: []}
+ for j, camn in enumerate(range(1, nb_cam + 1)):
+ uv = calib.find2d(1, dlt_coefs[j], np.array([[self.hybrid_skeleton3d[obj_name][label]['x'][i],
+ self.hybrid_skeleton3d[obj_name][label]['y'][i], self.hybrid_skeleton3d[obj_name][label]['z'][i]]]))
+ hybrid_skeleton3d_repro2d[label][camn] = np.array([uv[0][0], uv[0][1], 1.0])
+
+ if '3d' in res_method:
+ hybrid_skeleton2d = reproject_skeleton3d_to2d(hybrid_skeleton3d, dlt_coefs)
+
+ # Plot 2d repojection of hybrid skeleton
+ for j, camn in enumerate(range(1, nb_cam +1)):
+ center_of_mass_xy = \
+ (hybrid_skeleton2d_new['torso_abdomen_joint'][camn] * hybrid_param_ests_frame['ratio_com_torso'] \
+ + hybrid_skeleton2d_new['head_torso_joint'][camn] * (1 - hybrid_param_ests_frame['ratio_com_torso']))
+
+ if np.isnan(center_of_mass_xy).any(): continue
+
+ axs[j].clear()
+ axs[j].imshow(Image.open(os.path.join(rec_paths[camn], all_images[camn][i])))
+ for label in hybrid_skeleton2d_new.keys():
+ if hybrid_skeleton2d[label][camn][2] > self.hybrid_sets.threshold_likelihood:
+ axs[j].scatter(hybrid_skeleton2d[label][camn][0],
+ hybrid_skeleton2d[label][camn][1], marker='.', color='k')
+ axs[j].scatter(hybrid_skeleton3d_repro2d[label][camn][0],
+ hybrid_skeleton3d_repro2d[label][camn][1], marker='o', facecolors='none', edgecolors='k')
+ axs[j].scatter(hybrid_skeleton2d_init[label][camn][0],
+ hybrid_skeleton2d_init[label][camn][1], marker='o', facecolors='none', edgecolors='b')
+ axs[j].scatter(hybrid_skeleton2d_new[label][camn][0],
+ hybrid_skeleton2d_new[label][camn][1], marker='+', color='r')
+
+ axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
+ axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
+
+ plt.pause(0.05)
+ # plt.show()
+
+ # param_to_print = ['pitch_a', 'roll_a', 'yaw_a', 'ratio_hybrid']
+ # for label in param_to_print:
+ # print('{0}: init: {1} and est: {2}'.format(label, hybrid_params_init_frame[label], hybrid_param_ests_frame[label]))
+
+ else:
+ if '3d' in res_method:
+ args[i] = (self.hybrid_sets.skeleton3d_init, hybrid_skeleton3d, hybrid_params_init_frame, param_names,
+ res_method, opt_method, self.hybrid_bounds_init, return_nan,)
+
+ elif '2d' in res_method:
+ args[i] = (self.hybrid_sets.skeleton3d_init, hybrid_skeleton2d, hybrid_params_init_frame, param_names,
+ res_method, opt_method, self.hybrid_bounds_init, return_nan,)
+
+ if self.multiprocessing:
+ pool_hybrid = Pool(self.nb_process)
+ if '3d' in res_method:
+ results = pool_hybrid.starmap(optim_fit_body_params, args)
+
+ elif '2d' in res_method:
+ results = pool_hybrid.starmap(partial(optim_fit_body_params, dlt_coefs=dlt_coefs), args)
+
+ pool_hybrid.close()
+ for i, frame in enumerate(frames):
+ hybrid_param_ests_frame, rmse[obj_name][i], nb_iterations[obj_name][i] = results[i]
+ for label in hybrid_param_ests_frame.keys():
+ hybrid_param_ests[obj_name][label][i] = hybrid_param_ests_frame[label]
+
+ hybrid_param_ests[obj_name] = filter_interp_params(self.cutoff_frequency, self.frame_rate,
+ hybrid_param_ests[obj_name], param_names,
+ interpolate_nans, filter_high_freq, show_plot)
+
+ return hybrid_param_ests, nb_visible_pts, rmse, nb_iterations
+
+ def correct_body_wings_params(self, body_params, wings_params, angle_names_to_correct, rec_paths, dlt_coefs, show_plot=False):
+ # Will correct asymetrical wing angles (mean value should be the same on both side)
+ body_params_corrected = copy.deepcopy(body_params)
+ wings_params_corrected = copy.deepcopy(wings_params)
+
+ obj_names = self.skeleton2d['obj_names']
+ for obj_name in obj_names:
+
+ wings_params_mean = {} # Find angles error (btw stroke_a and deviation_a of both sides)
+ for side in self.wings_skeleton3d_init.keys():
+ wings_params_mean[side] = \
+ filter_interp_params(self.cutoff_frequency, self.frame_rate, wings_params[obj_name][side],
+ angle_names_to_correct, False, True, show_plot)
+
+ if 'stroke_a' in angle_names_to_correct:
+ diff_stroke_a = wings_params_mean['right']['stroke_a'] - wings_params_mean['left']['stroke_a']
+ # body_params_corrected[obj_name]['yaw_a'] -= diff_stroke_a / 2
+ # wings_params_corrected[obj_name]['right']['stroke_a'] -= diff_stroke_a / 2
+ # wings_params_corrected[obj_name]['left']['stroke_a'] += diff_stroke_a / 2
+ # wings_params_corrected[obj_name]['right']['yaw_a'] -= diff_stroke_a / 2
+ # wings_params_corrected[obj_name]['left']['yaw_a'] -= diff_stroke_a / 2
+
+ if 'deviation_a' in angle_names_to_correct:
+ diff_deviation_a = wings_params_mean['right']['deviation_a'] - wings_params_mean['left']['deviation_a']
+ # body_params_corrected[obj_name]['roll_a'] -= diff_deviation_a / 2
+ # wings_params_corrected[obj_name]['right']['deviation_a'] -= diff_deviation_a / 2
+ # wings_params_corrected[obj_name]['left']['deviation_a'] += diff_deviation_a / 2
+ # wings_params_corrected[obj_name]['right']['roll_a'] -= diff_deviation_a / 2
+ # wings_params_corrected[obj_name]['left']['roll_a'] -= diff_deviation_a / 2
+
+ if 'rotation_a' in angle_names_to_correct:
+ raise ValueError('There is not reason for rotation_a to be corrected!'
+ 'Any asymmetry btw sides cannot results from estimation error of the pitch angle')
+
+ if show_plot:
+ # Plot wing tips pos
+ frames = self.skeleton2d[obj_name]['frames']
+ wing_tips, wing_hinges = {'right': {}, 'left': {}}, {'right': {}, 'left': {}}
+ for side in self.wings_skeleton3d_init.keys():
+ for coord in ['x', 'y', 'z']:
+ wing_tips[side][coord], wing_hinges[side][coord] = np.ones(np.size(frames)) * np.nan, np.ones(
+ np.size(frames)) * np.nan
+
+ tip_hinge_a = np.ones(np.size(frames)) * np.nan
+ tip_torso_a = np.ones(np.size(frames)) * np.nan
+ for i, frame in enumerate(frames):
+ body_params_frame, wings_params_frame = {}, {'right': {}, 'left': {}}
+ for label in body_params[obj_name].keys():
+ body_params_frame[label] = body_params[obj_name][label][i]
+
+ body_skeleton3d_est = copy.deepcopy(self.body_skeleton3d_init)
+ body_skeleton3d_est = body.build_skeleton3d(body_skeleton3d_est, body_params_frame)
+ body_skeleton3d_est = body.rotate_skeleton3d(body_skeleton3d_est, body_params_frame)
+ body_skeleton3d_est = body.translate_skeleton3d(body_skeleton3d_est, body_params_frame)
+
+ wings_skeleton3d_mean = copy.deepcopy(self.wings_skeleton3d_init)
+ for side in self.wings_skeleton3d_init.keys():
+ for label in wings_params[obj_name][side].keys():
+ wings_params_frame[side][label] = wings_params_mean[side][label][i]
+
+ wings_skeleton3d_mean[side] = wings.build_skeleton3d(wings_skeleton3d_mean[side],
+ wings_params_frame[side])
+ wings_skeleton3d_mean[side] = wings.rotate_and_translate_skeleton3d(wings_skeleton3d_mean[side],
+ wings_params_frame[side],
+ side)
+
+ wing_tips[side]['x'][i], wing_tips[side]['y'][i], wing_tips[side]['z'][i] = \
+ wings_skeleton3d_mean[side]['tip'][0], wings_skeleton3d_mean[side]['tip'][1], \
+ wings_skeleton3d_mean[side]['tip'][2]
+ wing_hinges[side]['x'][i], wing_hinges[side]['y'][i], wing_hinges[side]['z'][i] = \
+ wings_skeleton3d_mean[side]['hinge'][0], wings_skeleton3d_mean[side]['hinge'][1], \
+ wings_skeleton3d_mean[side]['hinge'][2]
+
+ tip_vect = np.array(
+ [wing_tips['right']['x'][i], wing_tips['right']['y'][i], wing_tips['right']['z'][i]]) \
+ - np.array(
+ [wing_tips['left']['x'][i], wing_tips['left']['y'][i], wing_tips['left']['z'][i]])
+ hinge_vect = np.array(
+ [wing_hinges['right']['x'][i], wing_hinges['right']['y'][i], wing_hinges['right']['z'][i]]) \
+ - np.array(
+ [wing_hinges['left']['x'][i], wing_hinges['left']['y'][i], wing_hinges['left']['z'][i]])
+
+ torso_vect = body_skeleton3d_est['head_torso_joint'] - body_skeleton3d_est[
+ 'torso_abdomen_joint']
+
+ tip_hinge_a[i] = angle_from_vectors(tip_vect, hinge_vect)
+ tip_torso_a[i] = angle_from_vectors(tip_vect, torso_vect)
+
+ fig_test, axs_test = plt.subplots(1, 3)
+ fig_test.suptitle('Average position of wing tip over time')
+ for side in self.wings_skeleton3d_init.keys():
+ axs_test[0].plot(wing_tips[side]['x'] - body_params[obj_name]['x_com'], label=side + '_tip')
+ axs_test[1].plot(wing_tips[side]['y'] - body_params[obj_name]['y_com'], label=side + '_tip')
+ axs_test[2].plot(wing_tips[side]['z'] - body_params[obj_name]['z_com'], label=side + '_tip')
+
+ axs_test[0].plot(wing_hinges[side]['x'] - body_params[obj_name]['x_com'], label=side + '_hinge')
+ axs_test[1].plot(wing_hinges[side]['y'] - body_params[obj_name]['y_com'], label=side + '_hinge')
+ axs_test[2].plot(wing_hinges[side]['z'] - body_params[obj_name]['z_com'], label=side + '_hinge')
+
+ axs_test[0].set_ylabel('x (m)'), axs_test[1].set_ylabel('y (m)'), axs_test[2].set_ylabel('z (m)')
+
+ axs_test[2].set_xlabel('frames')
+ fig_test.legend()
+
+ fig_test2, axs_test2 = plt.subplots(1, 2)
+ fig_test2.suptitle('Average angle btw wing tip vector and torso vector over time')
+ axs_test2[0].plot(tip_hinge_a, label='tip_hinge_a')
+ axs_test2[0].plot(diff_stroke_a, label='diff_stroke_a')
+ axs_test2[0].plot(diff_deviation_a, label='diff_deviation_a')
+ # axs_test2[0].plot(diff_rotation_a, label='diff_rotation_a')
+ axs_test2[1].plot(tip_torso_a, label='tip_torso_a')
+ axs_test2[0].set_ylabel('angle (deg)')
+
+ fig_test2.legend()
+
+ self.plot_skeleton2d(body_params_corrected, wings_params_corrected, rec_paths, dlt_coefs)
+
+ return body_params_corrected, wings_params_corrected
+
+ def plot_skeleton2d(self, body_params, wings_params, rec_paths, dlt_coefs, modulo=1, save_images=False,
+ save_paths={}, show_plot=False, show_body=True, show_wings=True):
+
+ # TODO remove the folowing data in code
+ camns_to_rotate = [2]
+ degrees = 270
+ main_camn = 1
+
+ nb_cam = len(rec_paths.keys())
+
+ obj_names = self.skeleton2d['obj_names']
+ for obj_name in obj_names:
+ if not show_body and not show_wings: break
+
+ if show_plot:
+ fig, axs = plt.subplots(1, nb_cam, figsize=(15, 20))
+ plt.axis('off')
+
+ all_images = {}
+ for camn in range(1, nb_cam + 1):
+ all_files = sorted(os.listdir(rec_paths[camn]), key=lambda s: s[s.rfind('.') - nb_leading_zeros:s.rfind('.')])
+ all_images[camn] = [s for s in all_files if '.{0}'.format(image_format) in s]
+
+ frames = self.skeleton2d[obj_name]['frames']
+ for i, frame in enumerate(frames):
+ if not frame % modulo == 0.0 and not frame == 1: continue
+
+ body_skeleton2d_dlc = {}
+ for body_label in self.skeleton2d[obj_name]['label_names']:
+ body_skeleton2d_dlc[body_label] = {1: np.array([])}
+ for camn in range(1, nb_cam + 1):
+ if self.skeleton2d[obj_name][body_label][camn]['likelihood'][i] > self.body_sets.threshold_likelihood:
+ body_skeleton2d_dlc[body_label][camn] = \
+ np.array([self.skeleton2d[obj_name][body_label][camn]['x'][i],
+ self.skeleton2d[obj_name][body_label][camn]['y'][i],
+ self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
+ else:
+ body_skeleton2d_dlc[body_label][camn] = \
+ np.array([np.nan, np.nan, self.skeleton2d[obj_name][body_label][camn]['likelihood'][i]])
+
+ if show_body:
+ body_params_frame = {}
+ for label in body_params[obj_name].keys():
+ body_params_frame[label] = body_params[obj_name][label][i]
+
+ body_skeleton3d = copy.deepcopy(self.body_skeleton3d_init)
+ body_skeleton3d = body.build_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton3d = body.rotate_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton3d = body.translate_skeleton3d(body_skeleton3d, body_params_frame)
+ body_skeleton2d = reproject_skeleton3d_to2d(body_skeleton3d, dlt_coefs)
+
+ if show_wings:
+ wings_skeleton3d = copy.deepcopy(self.wings_skeleton3d_init)
+
+ wings_params_frame = {'right': {}, 'left': {}}
+ wings_skeleton2d, wings_skeleton2d_dlc = {'right': {}, 'left': {}}, {'right': {}, 'left': {}}
+ for side in self.wings_skeleton3d_init.keys():
+ for label in wings_params[obj_name][side].keys():
+ wings_params_frame[side][label] = wings_params[obj_name][side][label][i]
+
+ wings_skeleton3d[side] = wings.build_skeleton3d(wings_skeleton3d[side], wings_params_frame[side])
+ wings_skeleton3d[side] = wings.rotate_and_translate_skeleton3d(wings_skeleton3d[side], wings_params_frame[side], side)
+ wings_skeleton2d[side] = reproject_skeleton3d_to2d(wings_skeleton3d[side], dlt_coefs)
+
+ for wing_label in self.wings_skeleton3d_init[side].keys():
+ wings_skeleton2d_dlc[side][wing_label] = {1: []}
+ for camn in range(1, nb_cam + 1):
+ label = side + '_wing_' + wing_label
+ if self.skeleton2d[obj_name][label][camn]['likelihood'][i] > self.wings_sets.threshold_likelihood:
+ wings_skeleton2d_dlc[side][wing_label][camn] = \
+ np.array([self.skeleton2d[obj_name][label][camn]['x'][i],
+ self.skeleton2d[obj_name][label][camn]['y'][i],
+ self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
+ else:
+ wings_skeleton2d_dlc[side][wing_label][camn] = \
+ np.array([np.nan, np.nan, self.skeleton2d[obj_name][label][camn]['likelihood'][i]])
+
+ prev_img = None
+ for j, camn in enumerate(range(1, nb_cam + 1)):
+ center_of_mass_xy = \
+ (body_skeleton2d['torso_abdomen_joint'][camn] * body_params_frame['ratio_com_torso'] \
+ + body_skeleton2d['head_torso_joint'][camn] * (1 - body_params_frame['ratio_com_torso']))
+
+ if np.isnan(center_of_mass_xy).any(): continue
+
+ if show_plot or save_images:
+ img = read_image(os.path.join(rec_paths[camn], all_images[camn][i]))
+
+ if show_plot:
+ axs[j].clear()
+ axs[j].imshow(img)
+
+ if show_body:
+ for label in body_skeleton2d.keys():
+ if body_skeleton2d_dlc[label][camn][2] > self.body_sets.threshold_likelihood:
+ axs[j].scatter(body_skeleton2d_dlc[label][camn][0],
+ body_skeleton2d_dlc[label][camn][1], marker='.', color='k')
+ axs[j].scatter(body_skeleton2d[label][camn][0],
+ body_skeleton2d[label][camn][1], marker='+', color='r')
+
+ if show_wings:
+ for side in self.wings_skeleton3d_init.keys():
+ for label in wings_skeleton2d[side].keys():
+ if wings_skeleton2d_dlc[side][label][camn][2] > self.wings_sets.threshold_likelihood:
+ axs[j].scatter(wings_skeleton2d_dlc[side][label][camn][0],
+ wings_skeleton2d_dlc[side][label][camn][1], marker='.', color='k')
+ axs[j].scatter(wings_skeleton2d[side][label][camn][0],
+ wings_skeleton2d[side][label][camn][1], marker='+', color='r')
+
+ axs[j].set_xlim((center_of_mass_xy[0] - 100, center_of_mass_xy[0] + 100))
+ axs[j].set_ylim((center_of_mass_xy[1] + 100, center_of_mass_xy[1] - 100))
+
+ if camn == nb_cam:
+ plt.pause(0.05)
+ # fig.savefig(os.path.join(save_paths[main_camn], all_images[camn][i]),
+ # bbox_inches='tight', transparent=True, pad_inches=0)
+
+ if save_images:
+ img = img.convert('RGB')
+ draw = ImageDraw.Draw(img)
+ if show_body:
+ body_line = []
+ for label in body_skeleton2d.keys():
+ if body_skeleton2d_dlc[label][camn][2] > self.body_sets.threshold_likelihood:
+ draw.point((body_skeleton2d_dlc[label][camn][0], body_skeleton2d_dlc[label][camn][1]), fill='blue')
+ draw.point((body_skeleton2d[label][camn][0], body_skeleton2d[label][camn][1]), fill='red')
+
+ if not 'hinge' in label:
+ body_line.append((body_skeleton2d[label][camn][0], body_skeleton2d[label][camn][1]))
+ draw.line(body_line, fill='red', width=2)
+
+ if show_wings:
+ for side in self.wings_skeleton3d_init.keys():
+ wing_line = []
+ for label in wings_skeleton2d[side].keys():
+ if wings_skeleton2d_dlc[side][label][camn][2] > self.wings_sets.threshold_likelihood:
+ draw.point((wings_skeleton2d_dlc[side][label][camn][0], wings_skeleton2d_dlc[side][label][camn][1]), fill='blue')
+
+ draw.point((wings_skeleton2d[side][label][camn][0], wings_skeleton2d[side][label][camn][1]), fill='red')
+ wing_line.append((wings_skeleton2d[side][label][camn][0], wings_skeleton2d[side][label][camn][1]))
+
+ frst_label = list(wings_skeleton2d[side].keys())[0]
+ wing_line.append((wings_skeleton2d[side][frst_label][camn][0], wings_skeleton2d[side][frst_label][camn][1]))
+ draw.line(wing_line, fill='red', width=2)
+
+ left, right = int(center_of_mass_xy[0] - 100), int(center_of_mass_xy[0] + 100)
+ top, bottom = int(center_of_mass_xy[1] - 100), int(center_of_mass_xy[1] + 100)
+ img = img.crop((left, top, right, bottom))
+
+ if camn in camns_to_rotate:
+ img = img.rotate(degrees)
+ # img.show()
+ # img.save(os.path.join(save_paths[camn], all_images[camn][i]), compression='lzw')
+ # img.close()
+
+ # Stitch various cam views
+ if prev_img is None:
+ prev_img = img.copy()
+
+ else:
+ dst = Image.new('RGB', (prev_img.width + img.width, prev_img.height))
+ dst.paste(prev_img, (0, 0))
+ dst.paste(img, (prev_img.width, 0))
+ prev_img = dst.copy()
+
+ if not prev_img is None:
+ save_name = all_images[main_camn][i][:-len(image_format) - 1] \
+ + '-' + obj_name + '.{0}'.format(image_format)
+ prev_img.save(os.path.join(save_paths[main_camn], save_name), compression='lzw')
+ prev_img.close()
+ img.close()
+
+
+def filter_interp(y, frame_rate, cutoff_frequency):
+ y_filtered = y.copy()
+ is_nan, x = np.isnan(y), lambda z: z.nonzero()[0]
+
+ if sum(~is_nan) == 0: return y_filtered # only nans
+ if sum(is_nan) > 0: # interpolation needed for filter_high_freq
+ y_filtered[is_nan] = np.interp(x(is_nan), x(~is_nan), y_filtered[~is_nan])
+
+ w = cutoff_frequency / (frame_rate / 2) # Normalize the frequency
+ b, a = signal.butter(2, w, 'low')
+
+ y_filtered = signal.filtfilt(b, a, y_filtered)
+ y_filtered[is_nan] = np.nan
+
+ return y_filtered
+
+
+def minimize_perp_distance(x, y, z):
+ def model(params, xyz):
+ a, b, c, d = params
+ x, y, z = xyz
+ length_squared = a ** 2 + b ** 2 + c ** 2
+ return ((a * x + b * y + c * z + d) ** 2 / length_squared).sum()
+
+ def unit_length(params):
+ a, b, c, d = params
+ return a ** 2 + b ** 2 + c ** 2 - 1
+
+ # Initial guess of params using first three points
+ p1 = np.squeeze(np.transpose([x[0], y[0], z[0]]))
+ p2 = np.squeeze(np.transpose([x[1], y[1], z[1]]))
+ p3 = np.squeeze(np.transpose([x[2], y[2], z[2]]))
+
+ v1, v2 = p3 - p1, p2 - p1 # These two vectors are in the plane
+ cp = np.cross(v1, v2) # the cross product is a vector normal to the plane
+ d = np.dot(cp, p3) # This evaluates a * x3 + b * y3 + c * z3 which equals d
+ a, b, c = cp
+
+ # constrain the vector perpendicular to the plane be of unit length
+ cons = ({'type': 'eq', 'fun': unit_length})
+ sol = optimize.minimize(model, (a, b, c, d), args=[x, y, z], constraints=cons)
+
+ return tuple(sol.x)
\ No newline at end of file
diff --git a/process/smoothing_filtering.py b/process/smoothing_filtering.py
new file mode 100644
index 0000000000000000000000000000000000000000..73ecb606ac89f9a649f4872cb1c99bd369c9a4fd
--- /dev/null
+++ b/process/smoothing_filtering.py
@@ -0,0 +1,57 @@
+import copy
+
+import numpy as np
+import matplotlib.pyplot as plt
+
+from scipy import signal
+
+
+def filter_interp_params(cutoff_frequency, frame_rate, params, param_names, interpolate_nans, filter_high_freq,
+ show_plot=False):
+ if not interpolate_nans and not filter_high_freq: return params
+ new_params = copy.deepcopy(params)
+
+ if show_plot: fig, axs = plt.subplots(len(param_names), 1)
+
+ for i, label in enumerate(param_names):
+ is_nan, x = np.isnan(new_params[label]), lambda z: z.nonzero()[0]
+ if sum(~is_nan) == 0: continue # only nans
+
+ if show_plot: axs[i].plot(new_params[label], label='raw')
+
+ if '_a' in label:
+ # fig2 = plt.figure()
+ # ax2 = fig2.add_subplot(111)
+ # ax2.plot(new_params[label], label='raw')
+
+ new_params[label][~is_nan] = np.unwrap(np.deg2rad(new_params[label][~is_nan]))
+ # ax2.plot(np.rad2deg(new_params[label]), label='unwrapped')
+
+ if sum(is_nan) > 0: # interpolation needed for filter_high_freq
+ new_params[label][is_nan] = np.interp(x(is_nan), x(~is_nan), new_params[label][~is_nan])
+
+ if filter_high_freq and np.nanstd(new_params[label]) > 10e-6:
+ w = cutoff_frequency / (frame_rate / 2) # Normalize the frequency
+ b, a = signal.butter(2, w, 'low')
+
+ new_params[label] = signal.filtfilt(b, a, new_params[label])
+
+ if not interpolate_nans and sum(is_nan) > 0: # to fill back with nan if didnt want to interpolate
+ new_params[label][is_nan] = np.nan
+
+ if '_a' in label: # Wrap angles on [-pi, pi)
+ new_params[label] = np.rad2deg((new_params[label] + np.pi) % (2 * np.pi) - np.pi)
+ # ax2.plot(new_params[label], label='wrapped')
+ # ax2.set_ylabel(label)
+ # ax2.set_xlabel('frames')
+
+ if show_plot:
+ axs[i].plot(new_params[label], label='filt')
+ axs[i].set_ylabel(label)
+ axs[i].set_xlabel('frames')
+
+ if show_plot:
+ plt.legend()
+ # plt.show()
+
+ return new_params
\ No newline at end of file
diff --git a/process/utils.py b/process/utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..4241fe065841be7ef9d650ed45dc822933b84876
--- /dev/null
+++ b/process/utils.py
@@ -0,0 +1,296 @@
+import os, re, copy
+
+import inspect
+from typing import List, Dict
+
+from difflib import SequenceMatcher
+from datetime import datetime
+
+import numpy as np
+
+
+def match_folders_by_name(folder_names: List[str],
+ threshold_ratio_diff: float = None,
+ threshold_nb_diff_char: int = None) -> List[List[str]]:
+ """
+ Matches list of folder names together by checking how similar their folder names are and associate them together
+ The best matches will be found by comparing to the smaller list of folders to avoid matching one to two folders.
+
+ Args:
+ folder_names: List[list[]] of folder names
+ threshold_ratio_diff: Will check the similarity percentage (between 0 and 1)
+ threshold_nb_diff_char: Will check how many characters are different between the file names.
+
+ Returns:
+ matched_folder_names: List of lists (# of recordings * # of cameras) containing the matched folder names.
+
+ Raises:
+ ValueError: If both threshold type are given instead of only one
+ TypeError: thresholds have to be either a positive int or a float between 0.0 and 1.0.
+ """
+
+ if threshold_ratio_diff is not None and threshold_nb_diff_char is not None:
+ raise ValueError("Only one threshold has to be procured")
+
+ elif threshold_ratio_diff is None and threshold_nb_diff_char is None:
+ threshold_ratio_diff = 1.0
+
+ if threshold_ratio_diff is not None:
+ assert type(threshold_ratio_diff) is float
+ assert 0.0 <= threshold_ratio_diff <= 1.0
+
+ elif threshold_nb_diff_char is not None:
+ if type(threshold_nb_diff_char) is float and threshold_nb_diff_char.is_integer():
+ threshold_nb_diff_char = int(threshold_nb_diff_char)
+ assert threshold_nb_diff_char >= 0
+
+ else:
+ raise TypeError('Threshold has to be either a positive int or a float between 0.0 and 1.0')
+
+ nb_directories = len(folder_names)
+ nb_folders = [len(x) for x in folder_names]
+ i_min = np.argmin(nb_folders)
+
+ matched_folder_names, ratios = [[]] * len(folder_names), [[]] * len(folder_names)
+ is_similar_enough = [[]] * len(folder_names)
+ for i in range(nb_directories):
+
+ if i == i_min:
+ matched_folder_names[i] = folder_names[i_min]
+ ratios[i] = [1.0] * len(folder_names[i_min])
+
+ else:
+ matches = [max([(SequenceMatcher(a=f1, b=f2).ratio(), f2) for f2 in folder_names[i]]) for f1 in folder_names[i_min]]
+
+ matched_folder_names[i] = [folder_name for ratio, folder_name in matches]
+ ratios[i] = [ratio for ratio, folder_name in matches]
+
+ if threshold_ratio_diff is not None:
+ is_similar_enough[i] = [ratio >= threshold_ratio_diff for ratio in ratios[i]]
+
+ elif threshold_nb_diff_char is not None:
+ nb_char_diffs = [round((1-ratio)*len(folder_names[i_min][j])) for j, ratio in enumerate(ratios[i])]
+ is_similar_enough[i] = [nb_char_diff <= threshold_nb_diff_char for nb_char_diff in nb_char_diffs]
+
+ matched_folder_names = [list(x) for x in zip(*matched_folder_names)]
+ is_similar_enough = [list(x) for x in zip(*is_similar_enough)]
+
+ matched_folder_names = [x for i, x in enumerate(matched_folder_names) if all(is_similar_enough[i])]
+
+ return matched_folder_names
+
+
+def match_recordings_by_name(paths: List[str],
+ threshold_ratio_diff: float = None,
+ threshold_nb_diff_char: int = None) -> List[List[str]]:
+ """
+ Matches the recordings together (minimum 2) by checking how similar their folder names are and associate them together
+ The best matches will be found by comparing to the smaller list of folders to avoid matching one to two other folders.
+
+ Args:
+ paths: List of directory paths (one per camera) each containing folders (to be matched) with recorded images
+ threshold_ratio_diff: Will check the similarity percentage (between 0 and 1)
+ threshold_nb_diff_char: Will check how many characters are different between the file names.
+
+ Returns:
+ matched_folder_names: List of lists (# of recordings * # of cameras) containing the matched folder names.
+ """
+
+ folder_names = get_list_of_folder_names(paths)
+
+ if threshold_nb_diff_char is not None:
+ return match_folders_by_name(folder_names, threshold_nb_diff_char=threshold_nb_diff_char)
+
+ else:
+ return match_folders_by_name(folder_names, threshold_ratio_diff=threshold_ratio_diff)
+
+
+def match_folders_by_date(folder_names: List[str],
+ expr: str = r"\d{8}_\d{6}",
+ format_date_str: str ='%Y%m%d_%H%M%S',
+ threshold_s: int = 30) -> List[List[str]]:
+ """
+ Matches list of folder names together by checking how similar their folder names are and associate them together
+ The best matches will be found by comparing to the smaller list of folders to avoid matching one to two folders.
+
+ Args:
+ folder_names: List[list[]] of folder names
+ expr: Regular expression pattern matching recording folders.
+ format_date_str: format of the date and time in the folder names.
+ threshold_s: Will check if the date in the folder names differ from less than threshold in second
+
+ Returns:
+ matched_folder_names: List of lists (# of recordings * # of cameras) containing the matched folder names.
+
+ Raises:
+ TypeError: threshold_s has to be a positive int or a float between 0.0 and 1.0.
+ """
+
+ if type(threshold_s) is float and threshold_s.is_integer():
+ threshold_s = int(threshold_s)
+
+ if not type(threshold_s) is int and threshold_s >= 0:
+ raise TypeError('Threshold has to be either a positive int or a float between 0.0 and 1.0')
+
+ pattern = re.compile(expr)
+
+ dates = [[]] * len(folder_names)
+ for i in range(len(folder_names)):
+
+ new_folder_names, new_dates = [], []
+ for j in range(len(folder_names[i])):
+ match = re.search(pattern, folder_names[i][j])
+
+ if match:
+ new_folder_names.append(folder_names[i][j])
+ new_dates.append(datetime.strptime(match.group(), format_date_str))
+
+ folder_names[i] = new_folder_names
+ dates[i] = new_dates
+
+ nb_folders = [len(x) for x in folder_names]
+ i_min = np.argmin(nb_folders)
+
+ matched_folder_names, diff_time_s = [[]] * len(folder_names), [[]] * len(folder_names)
+ is_similar_enough = [[]] * len(folder_names)
+ for i in range(len(folder_names)):
+
+ if i == i_min:
+ matched_folder_names[i] = folder_names[i_min]
+ diff_time_s[i] = [0] * len(folder_names[i_min])
+
+ else:
+ matches = [min([(abs((d1 - d2).total_seconds()), folder_names[i][j])
+ for j, d2 in enumerate(dates[i])]) for d1 in dates[i_min]]
+ matched_folder_names[i] = [folder_name for diff_time_s, folder_name in matches]
+ diff_time_s[i] = [diff_s for diff_s, folder_name in matches]
+
+ is_similar_enough[i] = [x <= threshold_s for x in diff_time_s[i]]
+
+ matched_folder_names = [list(x) for x in zip(*matched_folder_names)]
+ is_similar_enough = [list(x) for x in zip(*is_similar_enough)]
+
+ matched_folder_names = [x for i, x in enumerate(matched_folder_names) if all(is_similar_enough[i])]
+
+ return matched_folder_names
+
+
+def match_recordings_by_date(paths: List[str],
+ expr: str = r"\d{8}_\d{6}",
+ format_date_str: str ='%Y%m%d_%H%M%S',
+ threshold_s: int = 30) -> List[List[str]]:
+ """
+ Matches the recordings together (minimum 2) by checking how similar the date in their folder names are
+ and associate them together. The best matches will be found by comparing to the smaller list of folders
+ to avoid matching one to two other folders.
+
+ Args:
+ paths: List of directory paths (one per camera) each containing folders (to be matched) with recorded images
+ expr: Regular expression pattern matching recording folders.
+ format_date_str: format of the date and time in the folder names.
+ threshold_s: Will check if the date in the folder names differ from less than threshold in second
+
+ Returns:
+ matched_folder_names: List of lists (# of recordings * # of cameras) containing the matched folder names.
+ """
+
+ folder_names = get_list_of_folder_names(paths)
+
+ return match_folders_by_date(folder_names, expr, format_date_str, threshold_s)
+
+
+def get_unmatched_folders(folder_names: List[List[str]],
+ matched_folder_names: List[List[str]]) -> List[List[str]]:
+
+ if len(folder_names) == len(matched_folder_names[0]):
+ matched_folder_names = [list(x) for x in zip(*matched_folder_names)]
+
+ unmatched_folder_names = [[]] * len(folder_names)
+ for i in range(len(folder_names)):
+
+ unmatched_folder_names[i] = [folder_names[i][j] for j in range(len(folder_names[i]))
+ if folder_names[i][j] not in matched_folder_names[i]]
+
+ return unmatched_folder_names
+
+
+def get_unmatched_recordings(paths: List[str],
+ matched_folder_names: List[List[str]]) -> List[List[str]]:
+
+ folder_names = get_list_of_folder_names(paths)
+
+ unmatched_folder_names = get_unmatched_folders(folder_names, matched_folder_names)
+
+ return unmatched_folder_names
+
+
+def get_list_of_folder_names(paths: List[str], keep_only_folder_with_images: bool = True) -> List[List[str]]:
+
+ folder_names = [[]] * len(paths)
+ for ind_cam, directory in enumerate(paths):
+ folders = os.listdir(directory)
+ folder_names[ind_cam] = [os.path.basename(os.path.normpath(folder)) for folder in folders]
+
+ if keep_only_folder_with_images:
+
+ ind_folder_to_remove = []
+ for ind_folder, folder_name in enumerate(folder_names[ind_cam]):
+ if not does_folder_contains_images(os.path.join(directory, folder_name)):
+ ind_folder_to_remove.append(ind_folder)
+
+ if ind_folder_to_remove:
+ [folder_names[ind_cam].pop(i) for i in ind_folder_to_remove[::-1]]
+
+ return folder_names
+
+
+def get_image_names_in_directory(directory: str, expr: str = r"\d+(?=\.(tif|png|jpg|bmp))") -> List[str]:
+ """
+
+ Args:
+ directory: The path to a directory containing images.
+ expr: Regular expression pattern matching image files.
+ """
+
+ if not os.path.exists(directory): return []
+
+ contents = os.listdir(directory)
+ pattern = re.compile(expr)
+
+ image_names = []
+ for file_name in contents:
+ match = re.search(pattern, file_name)
+ if match:
+ image_names.append(os.path.basename(os.path.normpath(file_name)))
+
+ return image_names
+
+
+def does_folder_contains_images(directory: str, expr: str = r"\d+(?=\.(tif|png|jpg|bmp))") -> bool:
+ """
+
+ Args:
+ directory: The path to a directory containing images.
+ expr: Regular expression pattern matching image files.
+ """
+
+ image_names = get_image_names_in_directory(directory, expr)
+
+ return len(image_names) != 0
+
+
+def convert_kwargs_to_be_writen_in_yaml(kwargs: Dict[any, any]) -> Dict[any, any]:
+
+ new_kwargs = copy.deepcopy(kwargs)
+ for key in kwargs.keys():
+ if hasattr(new_kwargs[key], '__dict__'):
+ new_kwargs[key] = new_kwargs[key].__dict__
+
+ elif type(new_kwargs[key]) is dict:
+ new_kwargs = convert_kwargs_to_be_writen_in_yaml(new_kwargs)
+
+ return new_kwargs
+
+
+def find_char(s, ch):
+ return [i for i, ltr in enumerate(s) if ltr == ch]
\ No newline at end of file
diff --git a/processes.yaml b/processes.yaml
index 60023f51a05ec7c33dbfcb6affb69b3c22b981f3..b8611176b2ed592fc3ed20cff78911b67ae2eb78 100644
--- a/processes.yaml
+++ b/processes.yaml
@@ -23,7 +23,7 @@
5:
fn: rotate
kwargs:
- camns:
+ camn_to_process:
- 2
degrees: 270
from_fn_name: crop
diff --git a/requirements.txt b/requirements.txt
index 83014c999fc427c516502497419ea97d1bd14615..d9095b26516e9a4923001ae18fb11f52f9c1ade3 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -8,6 +8,8 @@ scipy==1.4.1
pandas~=1.0.4
setuptools~=57.0.0
+pytest~=7.0.1
+
PyYAML~=3.13
pytest~=7.0.1
diff --git a/scripts/gen_dlt_coefs.py b/scripts/gen_dlt_coefs.py
index a70469965b469391e6c4965fa95791c5b64f9c27..bf23bfc40e7038e9ab32162038af41207da8f7ba 100644
--- a/scripts/gen_dlt_coefs.py
+++ b/scripts/gen_dlt_coefs.py
@@ -1,8 +1,8 @@
import sys, os
import numpy as np
-from cam import calib
+from camera import calib
-from point_tracker.reconstructor3d import gen_dlt
+from camera.calib import read_dlt_coefs
corePath = os.path.join("/home", "user", "Desktop", "John", "_Python", "dlc_flytracker", "camera")
sys.path.append(corePath)
@@ -57,14 +57,16 @@ dlt_path = os.path.join(calib_path, dlt_name)
# gen_dlt(3, img_size, xyz_path, xy_path, dlt_path, points_to_remove=[1, 3, 4, 11, 13, 20], plot_error_analysis=False) # small
# Load dlt_coefs
-dlt_coefs = np.zeros((nb_cam, 12))
+dlt_coefs = np.zeros((nb_cam, 11))
dlt_csv = np.genfromtxt(dlt_path, delimiter=',', skip_header=0, names=True)
dlt_csv.dtype.names = ['cam{0}'.format(camn) for camn in range(1, nb_cam + 1)]
for i, camn in enumerate(range(1, nb_cam + 1)):
dlt_coefs[i] = np.append(np.array(dlt_csv['cam{0}'.format(camn)]), 1)
+dlt_coefs = read_dlt_coefs(dlt_path)
+
# Reconstruct 3d and reproject to 2d
-obj_coord = calib.recon3D(3, nb_cam, dlt_coefs, pts_cams)
+obj_coord = calib.recon3d(3, nb_cam, dlt_coefs, pts_cams)
repro_pts_cams = calib.find2d(nb_cam, dlt_coefs, np.array([obj_coord]))
mean_repro_err = np.sqrt(np.mean(np.sum((repro_pts_cams - pts_cams) ** 2, 1)))
diff --git a/scripts/process_all_escapes.py b/scripts/process_all_escapes.py
index fb4aea8dff9f0e2256aa55d747cc76ea48e761c5..bcba8bb2256de221631d1ad833bdb7126119b6d6 100644
--- a/scripts/process_all_escapes.py
+++ b/scripts/process_all_escapes.py
@@ -1,5 +1,5 @@
import os, time, yaml
-from images.process import ImagesProcessing
+from process.batch_processing import BatchProcessing
from convert_xypts import *
@@ -21,7 +21,7 @@ wing_param_names = ['stroke_a', 'deviation_a', 'rotation_a']
img_size = (1024, 1024)
-img_process = ImagesProcessing(3, dlt_path)
+img_process = BatchProcessing()
img_process.cam_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1',
2: '/media/user/MosquitoEscape_Photron2/Photron2',
3: '/media/user/MosquitoEscape_Photron3/Photron3'}
@@ -81,27 +81,27 @@ for i, date in enumerate(dates_to_process):
# rec_names_to_process = non_processed_rec_names
# # Do cropping, rotating, stitching, etc at once and save as .avi (save much time and space)
- # img_process.do_batch('multi_processes', rec_names=rec_names_to_process, yaml_path='processes_updated.yaml', delete_previous=True)
+ # img_process.do_batch('multi_processes', rec_names_to_process=rec_names_to_process, yaml_path='processes_updated.yaml', delete_previous=True)
# # Do 2d tracking (blob detection) on images in cam_save_paths/_Sample
- # img_process.do_batch('track2d', rec_names=rec_names_to_process, from_fn_name='sample')
+ # img_process.do_batch('track2d', rec_names_to_process=rec_names_to_process, from_fn_name='sample')
#
# # Do 3d reconstruction of tracks
- # img_process.do_batch('recon3d', rec_names=rec_names_to_process, from_fn_name='sample')
+ # img_process.do_batch('recon3d', rec_names_to_process=rec_names_to_process, from_fn_name='sample', dlt_path=dlt_path)
# # Track features using DeepLabCut
- # img_process.do_batch('analyse_dlc', rec_names=rec_names_to_process, from_fn_name='save_avi', cfg_path=dlc_cfg_path,
- # shuffle=shuffle, trainingsetindex=0, batchsize=5, save_avi=False, model_name=model_name, delete_previous=True)
+ # img_process.do_batch('analyse_dlc', rec_names_to_process=rec_names_to_process, from_fn_name='save_avi', cfg_path=dlc_cfg_path,
+ # shuffle=shuffle, trainingset_index=0, batch_size=5, save_avi=False, model_name=model_name, delete_previous=True)
# # Load (+ filtering and unscrambling) 2d coords from DLC + Reverse processes (unstitch, rotate back, uncrop) + Reconstruct 3d coord
- # img_process.do_batch('load_dlc', rec_names=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
- # tracker_method='skeleton')
+ # img_process.do_batch('load_dlc', rec_names_to_process=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
+ # tracker_method='skeleton', dlt_path=dlt_path)
# Optimize fit of skeleton to find body and wings anglJues
- img_process.multiprocessing = True
+ img_process.do_multiprocessing = True
img_process.threshold_likelihood = 0.85
- img_process.do_batch('fit_skeleton', rec_names=rec_names_to_process, from_fn_name='load_dlc', model_name=model_name,
- csv_path=csv_wings_geo_path, body_param_names=body_param_names, wing_param_names=wing_param_names,
- animal_name="fly", res_method=res_method, opt_method=opt_method)
+ img_process._do_batch('fit_skeleton', rec_names_to_process=rec_names_to_process, from_fn_name='load_dlc', model_name=model_name,
+ csv_path=csv_wings_geo_path, body_param_names=body_param_names, wing_param_names=wing_param_names,
+ animal_name="fly", res_method=res_method, opt_method=opt_method, dlt_path=dlt_path)
print('> All processes as been processed (total time elapsed: {0:.4f} s)'.format(time.time() - start))
diff --git a/scripts/process_all_escapes_strobe.py b/scripts/process_all_escapes_strobe.py
index 66a59b77769e2c4aa0c2fee0db605deb714eee7c..35fc5b47397a866a5c5ba13d34895cba5eab1751 100644
--- a/scripts/process_all_escapes_strobe.py
+++ b/scripts/process_all_escapes_strobe.py
@@ -1,5 +1,5 @@
import time, yaml, os
-from images.process import ImagesProcessing
+from process.batch_processing import BatchProcessing
from convert_xypts import *
@@ -10,7 +10,7 @@ xyz_path = os.path.join(os.getcwd(), 'data/calib/xyz_calibration_device_big.csv'
csv_wings_geo_path = os.path.join(os.getcwd(), 'acoluzzii_wing.csv')
dlc_cfg_path = '/home/user/Desktop/Antoine/_maDLC/config.yaml'
-img_process = ImagesProcessing(3, dlt_path)
+img_process = BatchProcessing()
img_process.cam_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1',
2: '/media/user/MosquitoEscape_Photron2/Photron2',
3: '/media/user/MosquitoEscape_Photron3/Photron3'}
@@ -50,30 +50,30 @@ for i, date in enumerate(dates_to_process):
#rec_names_to_process = {1: ['cam1_20200303_030117'], 2: ['cam2_20200303_030120'], 3: ['cam3_20200303_030120']}
# # Do cropping, rotating, stitching, etc at once and save as .avi (save much time and space)
- # img_process.do_batch('multi_processes', rec_names=rec_names_to_process, yaml_path='processes-strobe.yaml', delete_previous=True)
- # img_process.do_batch('stitch', rec_names=rec_names_to_process, from_fn_name='save_strobe')
+ # img_process.do_batch('multi_processes', rec_names_to_process=rec_names_to_process, yaml_path='processes-strobe.yaml', delete_previous=True)
+ # img_process.do_batch('stitch', rec_names_to_process=rec_names_to_process, from_fn_name='save_strobe')
- # # Convert images to 8 bits and reduce framerate (save in cam_save_paths/_Sample)
- # img_process.do_batch('sample', rec_names=rec_names_to_process, step_frame=50, delete_previous=True)
+ # # Convert images to 8 bits and reduce frame_rate (save in cam_save_paths/_Sample)
+ # img_process.do_batch('sample', rec_names_to_process=rec_names_to_process, step_frame=50, delete_previous=True)
#
# # Do 2d tracking (blob detection) on images in cam_save_paths/_Sample
- # img_process.do_batch('track2d', rec_names=rec_names_to_process, from_fn_name='sample')
+ # img_process.do_batch('track2d', rec_names_to_process=rec_names_to_process, from_fn_name='sample')
#
# # Do 3d reconstruction of tracks
- # img_process.do_batch('recon3d', rec_names=rec_names_to_process, from_fn_name='sample', dlt_path=dlt_path)
+ # img_process.do_batch('recon3d', rec_names_to_process=rec_names_to_process, from_fn_name='sample', dlt_path=dlt_path)
# Generate and save stroboscopic images
- img_process.do_batch('save_strobe', rec_names=rec_names_to_process, from_fn_name='sample', step_frame=5, radius=60, shape='circle')
- # img_process.do_batch('save_strobe', rec_names=rec_names_to_process, from_fn_name='sample', step_frame=5, radius=60,
- # shape='strip_height', camns=[2])
- # img_process.do_batch('save_strobe', rec_names=rec_names_to_process, from_fn_name='sample', step_frame=5, radius=60,
- # shape='strip_width', camns=[1, 3])
+ img_process._do_batch('save_strobe', rec_names_to_process=rec_names_to_process, from_fn_name='sample', step_frame=5, radius=60, shape='circle')
+ # img_process.do_batch('save_strobe', rec_names_to_process=rec_names_to_process, from_fn_name='sample', step_frame=5, radius=60,
+ # shape='strip_height', camn_to_process=[2])
+ # img_process.do_batch('save_strobe', rec_names_to_process=rec_names_to_process, from_fn_name='sample', step_frame=5, radius=60,
+ # shape='strip_width', camn_to_process=[1, 3])
# Rotate view 2 (to always have piston coming from right side)
- img_process.do_batch('rotate', rec_names=rec_names_to_process, from_fn_name='save_strobe', camns=[2], degrees=270)
+ img_process._do_batch('rotate', rec_names_to_process=rec_names_to_process, from_fn_name='save_strobe', camn_to_process=[2], degrees=270)
# Stitch all views together and save in cam_save_paths/_Stitched
- img_process.do_batch('stitch', rec_names=rec_names_to_process, from_fn_name='save_strobe')
+ img_process._do_batch('stitch', rec_names_to_process=rec_names_to_process, from_fn_name='save_strobe')
print('> All processes as been processed (total time elapsed: {0:.4f} s)'.format(time.time() - start))
#exit()
\ No newline at end of file
diff --git a/scripts/process_mating_events.py b/scripts/process_mating_events.py
index 90127bd5a97855f28f28cb434a873a39a3332120..703018a1b9dd4652fe457d8e99c3c5f30a4a8a0a 100644
--- a/scripts/process_mating_events.py
+++ b/scripts/process_mating_events.py
@@ -1,7 +1,7 @@
-from images.process import ImagesProcessing
+from process.batch_processing import BatchProcessing
import numpy as np
-img_process = ImagesProcessing(3)
+img_process = BatchProcessing()
img_process.cam_paths = {1: 'F:\\Photron1', 2: 'Y:\\', 3: 'Z:\\'}
img_process.cam_save_paths = {1: 'F:\\Photron1\\_Process', 2: 'Y:\\_Process', 3: 'Z:\\_Process'}
@@ -23,10 +23,10 @@ if len(rec_cnt_be_found) > 0:
exit()
rec_to_delete = (list(set(img_process.all_folders_rec_cam[mating_events['camn'][0]]) - set(rec_mating_names)))
-img_process.do_batch('delete', rec_names=rec_names_to_process, to_del_names=rec_to_delete)
+img_process._do_batch('delete', rec_names_to_process=rec_names_to_process, to_del_names=rec_to_delete)
# Delete all without correspondings recording for camera 2 and 3
img_process.main_cam = 2
-img_process.do_batch('delete', rec_names=rec_names_to_process, to_del_names=[])
+img_process._do_batch('delete', rec_names_to_process=rec_names_to_process, to_del_names=[])
img_process.main_cam = 3
-img_process.do_batch('delete', rec_names=rec_names_to_process, to_del_names=[])
\ No newline at end of file
+img_process._do_batch('delete', rec_names_to_process=rec_names_to_process, to_del_names=[])
\ No newline at end of file
diff --git a/scripts/process_selection.py b/scripts/process_selection.py
index a375d2592edf6ae433928a246ee2895330e2c1cc..6246d3621d677e8b469d9420c7505ade93fb2e03 100644
--- a/scripts/process_selection.py
+++ b/scripts/process_selection.py
@@ -1,5 +1,5 @@
import os, time, yaml
-from images.process import ImagesProcessing
+from process.batch_processing import BatchProcessing
import numpy as np
selection_csv = 'selection_escapes.csv'
@@ -12,7 +12,7 @@ csv_wings_geo_path = os.path.join(os.getcwd(), '../acoluzzii_wing.csv')
dlc_cfg_path = '/home/user/Desktop/Antoine/_maDLC/config.yaml'
-img_process = ImagesProcessing(3, dlt_path)
+img_process = BatchProcessing()
img_process.cam_paths = {1: '/media/user/MosquitoEscape_Photron1/Photron1',
2: '/media/user/MosquitoEscape_Photron2/Photron2',
3: '/media/user/MosquitoEscape_Photron3/Photron3'}
@@ -89,55 +89,56 @@ for i, date in enumerate(dates_to_process):
if len(rec_names_to_process[camn]) == 0: continue
# # Do cropping, rotating, stitching, etc at once and save as .avi (save much time and space)
- # img_process.do_batch('multi_processes', rec_names=rec_names_to_process, yaml_path='config.yaml', delete_previous=True)
+ # img_process.do_batch('multi_processes', rec_names_to_process=rec_names_to_process, yaml_path='config.yaml', delete_previous=True)
- # # Convert images to 8 bits and reduce framerate (save in cam_save_paths/_Sample)
- # img_process.do_batch('sample', rec_names=rec_names_to_process, step_frame=50, delete_previous=True)
+ # # Convert images to 8 bits and reduce frame_rate (save in cam_save_paths/_Sample)
+ # img_process.do_batch('sample', rec_names_to_process=rec_names_to_process, step_frame=50, delete_previous=True)
# # Do 2d tracking (blob detection) on images in cam_save_paths/_Sample
- # img_process.do_batch('track2d', rec_names=rec_names_to_process, from_fn_name='sample')
+ # img_process.do_batch('track2d', rec_names_to_process=rec_names_to_process, from_fn_name='sample')
# # Do 3d reconstruction of tracks
- # img_process.do_batch('recon3d', rec_names=rec_names_to_process, from_fn_name='sample')
+ # img_process.do_batch('recon3d', rec_names_to_process=rec_names_to_process, from_fn_name='sample', dlt_path=dlt_path)
# # Crop all frames and save in cam_save_paths/_Cropped
- # img_process.do_batch('crop', rec_names=rec_names_to_process, from_fn_name='sample', height_crop=200, width_crop=200, delete_previous=True)
+ # img_process.do_batch('crop', rec_names_to_process=rec_names_to_process, from_fn_name='sample', height_crop=200, width_crop=200, delete_previous=True)
#
# # Rotate view 2 (to always have piston coming from right side)
- # img_process.do_batch('rotate', rec_names=rec_names_to_process, from_fn_name='crop', camns=[2], degrees=270)
+ # img_process.do_batch('rotate', rec_names_to_process=rec_names_to_process, from_fn_name='crop', camn_to_process=[2], degrees=270)
#
# # Upsizing of the images (increase resolution)
- # #img_process.do_batch('resize', rec_names=rec_names_to_process, resize_ratio=2, from_fn_name='crop', delete_previous=True)
+ # #img_process.do_batch('resize', rec_names_to_process=rec_names_to_process, resize_ratio=2, from_fn_name='crop', delete_previous=True)
#
# # Enhance images (increase sharpness and contrast)
- # # img_process.do_batch('enhance_sharpness', factor=2.0, rec_names=rec_names_to_process, from_fn_name='resize', delete_previous=True)
- # #img_process.do_batch('enhance_contrast', factor=1.5, rec_names=rec_names_to_process, from_fn_name='enhance_sharpness')
+ # # img_process.do_batch('enhance_sharpness', factor=2.0, rec_names_to_process=rec_names_to_process, from_fn_name='resize', delete_previous=True)
+ # #img_process.do_batch('enhance_contrast', factor=1.5, rec_names_to_process=rec_names_to_process, from_fn_name='enhance_sharpness')
#
# # Stitch all views together and save in cam_save_paths/_Stitched
- # img_process.do_batch('stitch', rec_names=rec_names_to_process, from_fn_name='crop', delete_previous=True)
+ # img_process.do_batch('stitch', rec_names_to_process=rec_names_to_process, from_fn_name='crop', delete_previous=True)
# # Save recordings in .avi
- # img_process.do_batch('save_avi', rec_names=rec_names_to_process, from_fn_name='stitch', lossless=False, delete_previous=True)
+ # img_process.do_batch('save_avi', rec_names_to_process=rec_names_to_process, from_fn_name='stitch', lossless=False, delete_previous=True)
# # Track features using DeepLabCut
- # img_process.do_batch('analyse_dlc', rec_names=rec_names_to_process, from_fn_name='save_avi', cfg_path=dlc_cfg_path,
- # shuffle=shuffle, trainingsetindex=0, batchsize=5, save_avi=False, model_name=model_name,
+ # img_process.do_batch('analyse_dlc', rec_names_to_process=rec_names_to_process, from_fn_name='save_avi', cfg_path=dlc_cfg_path,
+ # shuffle=shuffle, trainingset_index=0, batch_size=5, save_avi=False, model_name=model_name,
# delete_previous=True)
#
# # Load (+ filtering and unscrambling) 2d coords from DLC + Reverse processes (unstitch, rotate back, uncrop) + Reconstruct 3d coord
- # img_process.do_batch('load_dlc', rec_names=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
- # tracker_method='skeleton')
+ # img_process.do_batch('load_dlc', rec_names_to_process=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
+ # tracker_method='skeleton', dlt_path=dlt_path)
# Optimize fit of skeleton to find body and wings angles
- img_process.multiprocessing = True
+ img_process.do_multiprocessing = True
img_process.threshold_likelihood = 0.8
- img_process.do_batch('fit_skeleton', rec_names=rec_names_to_process, from_fn_name='load_dlc', model_name=model_name,
- csv_path=csv_wings_geo_path, body_param_names=body_param_names, wing_param_names=wing_param_names,
- animal_name="fly", res_method=res_method, opt_method=opt_method, angles_to_correct=['stroke_a', 'deviation_a'])
+ img_process._do_batch('fit_skeleton', rec_names_to_process=rec_names_to_process, from_fn_name='load_dlc', model_name=model_name,
+ csv_path=csv_wings_geo_path, body_param_names=body_param_names, wing_param_names=wing_param_names,
+ animal_name="fly", res_method=res_method, opt_method=opt_method, dlt_path=dlt_path,
+ angles_to_correct=['stroke_a', 'deviation_a'])
# Save .png of skeleton fit on raw images
- img_process.do_batch('plot_skeleton', rec_names=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
- csv_path=csv_wings_geo_path, save_avi=True)
+ img_process._do_batch('plot_skeleton', rec_names_to_process=rec_names_to_process, from_fn_name='analyse_dlc', model_name=model_name,
+ csv_path=csv_wings_geo_path, dlt_path=dlt_path, save_avi=True)
print('> All processes as been processed (total time elapsed: {0:.4f} s)'.format(time.time() - start))
diff --git a/analysis/plot/__init__.py b/skeleton_fitter/animals/__init__.py
similarity index 100%
rename from analysis/plot/__init__.py
rename to skeleton_fitter/animals/__init__.py
diff --git a/skeleton_fitter/cost_functions/fly.py b/skeleton_fitter/animals/fly.py
similarity index 61%
rename from skeleton_fitter/cost_functions/fly.py
rename to skeleton_fitter/animals/fly.py
index 094cd749dc9a40cd0831621959163f427baf000b..34606675f4888efb08a6c077b2c6befe6fbf0bcf 100644
--- a/skeleton_fitter/cost_functions/fly.py
+++ b/skeleton_fitter/animals/fly.py
@@ -6,6 +6,35 @@ from skeleton_fitter.skeleton_modules import insect_hybrid_wings_body_slim
from skeleton_fitter.skeleton_modules import insect_wings
+class AnimalSettings:
+ # TODO make class with animal characteristics
+
+ """
+ An animal consist of multiple modules (one body + an even number of limbs (right and left))
+ for example: a fly consist of one body: insect_body_slim, and multiple limbs: 2*insect_wing + 6*insect_leg
+
+ """
+
+ body_module_name = ['insect_body_slim']
+ limb_module_names = ['insect_wing', 'insect_leg']
+ limb_numbers = [2, 6]
+
+
+
+
+
+class AnimalHybridSettings:
+ # TODO make class with animal hybrid characteristics + explain why hybrid exist and when to use it
+
+ body_module_name = ['insect_hybrid_wings_body_slim']
+ limb_module_names = ['insect_wing', 'insect_leg']
+ limb_numbers = [2, 6]
+
+
+
+
+
+
def cost_function(param_values, param_names, init_skeleton3d, init_params, raw_skeleton, options):
"""
Cost function (sum of residuals) to fit body or wings skeleton (in init_skeleton3d) to points (in raw_skeleton)
@@ -19,7 +48,7 @@ def cost_function(param_values, param_names, init_skeleton3d, init_params, raw_s
options: dict with various optional parameters
Returns:
- sum_res: Sum of residuals when comparing init_skeleton3d (after being transformed with param_values) to raw_skeleton
+ sum_res: Sum of residuals when comparing init_skeleton3d (after transformed with param_values) to raw_skeleton
"""
new_params = copy.deepcopy(init_params)
for i, label in enumerate(param_names):
@@ -29,39 +58,39 @@ def cost_function(param_values, param_names, init_skeleton3d, init_params, raw_s
if 'body' in options['res_method'] or 'hybrid' in options['res_method']:
body_skeleton3d = copy.deepcopy(init_skeleton3d)
if 'body' in options['res_method']:
- body_skeleton3d = insect_body_slim.build_body_skeleton3d(body_skeleton3d, new_params)
+ body_skeleton3d = insect_body_slim.build_skeleton3d(body_skeleton3d, new_params)
elif 'hybrid' in options['res_method']:
- body_skeleton3d = insect_hybrid_wings_body_slim.build_hybrid_skeleton3d(body_skeleton3d, new_params)
+ body_skeleton3d = insect_hybrid_wings_body_slim.build_skeleton3d(body_skeleton3d, new_params)
- body_skeleton3d = insect_body_slim.rotate_body_skeleton3d(body_skeleton3d, new_params)
- body_skeleton3d = insect_body_slim.translate_body_skeleton3d(body_skeleton3d, new_params)
+ body_skeleton3d = insect_body_slim.rotate_skeleton3d(body_skeleton3d, new_params)
+ body_skeleton3d = insect_body_slim.translate_skeleton3d(body_skeleton3d, new_params)
if '3d' in options['res_method']:
- residuals = insect_body_slim.residuals_body3d(body_skeleton3d, raw_skeleton, options['default_residual'])
+ residuals = insect_body_slim.residuals3d(body_skeleton3d, raw_skeleton, options['default_residual'])
elif '2d' in options['res_method']:
- residuals = insect_body_slim.residuals_body2d(body_skeleton3d, raw_skeleton, options['default_residual'], options['dlt_coefs'])
+ residuals = insect_body_slim.residuals2d(body_skeleton3d, raw_skeleton, options['default_residual'], options['dlt_coefs'])
elif 'wing' in options['res_method']:
if 'geo' in options['res_method']:
wing_geometry3d = copy.deepcopy(init_skeleton3d)
- wing_geometry3d = insect_wings.build_wing_geometry3d(wing_geometry3d, new_params)
- wing_geometry3d = insect_wings.rotate_and_translate_wing_geometry3d(wing_geometry3d, new_params, options['side'])
+ wing_geometry3d = insect_wings.build_geometry3d(wing_geometry3d, new_params)
+ wing_geometry3d = insect_wings.rotate_and_translate_geometry3d(wing_geometry3d, new_params, options['side'])
if '3d' in options['res_method']:
- residuals = insect_wings.residuals_wing3d_geometry3d(wing_geometry3d, raw_skeleton, options['default_residual'])
+ residuals = insect_wings.residuals3d_geometry3d(wing_geometry3d, raw_skeleton, options['default_residual'])
elif '2d' in options['res_method']:
- residuals = insect_wings.residuals_wing2d_geometry3d(wing_geometry3d, raw_skeleton, options['default_residual'], options['dlt_coefs'])
+ residuals = insect_wings.residuals2d_geometry3d(wing_geometry3d, raw_skeleton, options['default_residual'], options['dlt_coefs'])
else:
wing_skeleton3d = copy.deepcopy(init_skeleton3d)
- wing_skeleton3d = insect_wings.build_wing_skeleton3d(wing_skeleton3d, new_params)
- wing_skeleton3d = insect_wings.rotate_and_translate_wing_skeleton3d(wing_skeleton3d, new_params, options['side'])
+ wing_skeleton3d = insect_wings.build_skeleton3d(wing_skeleton3d, new_params)
+ wing_skeleton3d = insect_wings.rotate_and_translate_skeleton3d(wing_skeleton3d, new_params, options['side'])
if '3d' in options['res_method']:
- residuals = insect_wings.residuals_wing3d(wing_skeleton3d, raw_skeleton, options['default_residual'])
+ residuals = insect_wings.residuals3d(wing_skeleton3d, raw_skeleton, options['default_residual'])
elif '2d' in options['res_method']:
- residuals = insect_wings.residuals_wing2d(wing_skeleton3d, raw_skeleton, options['default_residual'], options['dlt_coefs'])
+ residuals = insect_wings.residuals2d(wing_skeleton3d, raw_skeleton, options['default_residual'], options['dlt_coefs'])
else:
print('WARN: {0} is not a valid method!'.format(options['res_method']))
@@ -71,7 +100,7 @@ def cost_function(param_values, param_names, init_skeleton3d, init_params, raw_s
return np.sum(residuals ** 2)
-# Useless??
+# TODO check if following is useless
# def replace_outliers(data, m=2.):
# data = np.array(data)
#
@@ -88,10 +117,10 @@ if __name__ == '__main__':
csv_wings_geo_path = os.path.join(os.getcwd(), '../data/acoluzzii_wing.csv')
- body_params_init = insect_body_slim.init_body_params()
- wings_params_init = insect_wings.init_wings_params(body_params_init)
+ body_params_init = insect_body_slim.init_params()
+ wings_params_init = insect_wings.init_params(body_params_init)
- body_skeleton3d_init = insect_body_slim.init_body_skeleton3d(body_params_init)
+ body_skeleton3d_init = insect_body_slim.init_skeleton3d(body_params_init)
wings_skeleton3d_init, wings_geometry3d_init = insect_wings.init_wings_skeleton_geometry3d_from_csv(wings_params_init, csv_wings_geo_path)
body_params = copy.deepcopy(body_params_init)
@@ -100,13 +129,13 @@ if __name__ == '__main__':
body_params['roll_a'] = -5.0
[body_params['x_com'], body_params['y_com'], body_params['z_com']] = [0.0, 5.0, 3.0]
- wings_params = insect_wings.init_wings_params(body_params)
+ wings_params = insect_wings.init_params(body_params)
wings_params['right']['stroke_a'] = -55.0
wings_params['right']['deviation_a'] = 10.0
wings_params['right']['rotation_a'] = 60.0
- body_skeleton3d_1 = insect_body_slim.rotate_body_skeleton3d(body_skeleton3d_init, body_params)
- body_skeleton3d_1 = insect_body_slim.translate_body_skeleton3d(body_skeleton3d_1, body_params)
+ body_skeleton3d_1 = insect_body_slim.rotate_skeleton3d(body_skeleton3d_init, body_params)
+ body_skeleton3d_1 = insect_body_slim.translate_skeleton3d(body_skeleton3d_1, body_params)
#plot_body(body_skeleton3d_1)
wings_skeleton3d_1, wings_geometry1 = copy.deepcopy(wings_skeleton3d_init), copy.deepcopy(wings_geometry3d_init)
@@ -116,10 +145,10 @@ if __name__ == '__main__':
[wings_params[side]['x_com'], wings_params[side]['y_com'], wings_params[side]['z_com']] = \
[body_params['x_com'], body_params['y_com'], body_params['z_com']]
- wings_skeleton3d_1[side] = insect_wings.rotate_and_translate_wing_skeleton3d(wings_skeleton3d_init[side], wings_params[side], side)
- wings_geometry1[side] = insect_wings.rotate_and_translate_wing_geometry3d(wings_geometry3d_init[side], wings_params[side], side)
+ wings_skeleton3d_1[side] = insect_wings.rotate_and_translate_skeleton3d(wings_skeleton3d_init[side], wings_params[side], side)
+ wings_geometry1[side] = insect_wings.rotate_and_translate_geometry3d(wings_geometry3d_init[side], wings_params[side], side)
- insect_wings.plot_wings(wings_skeleton3d_1)
+ insect_wings.plot_multi(wings_skeleton3d_1)
insect_wings.plot_body_and_wings(body_skeleton3d_1, wings_skeleton3d_1)
# Estimate body and wings parameters
diff --git a/skeleton_fitter/cost_functions/__init__.py b/skeleton_fitter/cost_functions/__init__.py
deleted file mode 100644
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..0000000000000000000000000000000000000000
diff --git a/skeleton_fitter/skeleton_modules/insect_body_slim.py b/skeleton_fitter/skeleton_modules/insect_body_slim.py
index c935f7961d1ff324747d610845f1eaee95a75a2d..b1f641b3956e5f78ce00c60d568cbd81bdac92f0 100644
--- a/skeleton_fitter/skeleton_modules/insect_body_slim.py
+++ b/skeleton_fitter/skeleton_modules/insect_body_slim.py
@@ -1,14 +1,40 @@
import copy
import numpy as np
-import matplotlib.pyplot as plt
+from dataclasses import dataclass
+import matplotlib.pyplot as plt
from scipy.spatial.transform import Rotation as R
from skeleton_fitter.utils import reproject_skeleton3d_to2d, angle_from_vectors, length_from_vector, set_axes3d_equal
+import skeleton_fitter.skeleton_modules.insect_body_slim as body
+
+
+@dataclass
+class ModuleSettings:
+ threshold_likelihood = 0.9
+ threshold_nb_pts = 4
+
+ params_init = body.init_params()
+ bounds_init = body.init_bounds()
+
+ skeleton2d_init = body.init_skeleton2d(params_init)
+ skeleton3d_init = body.init_skeleton3d(params_init)
+
+ labels = list(skeleton3d_init.keys())
+ # labels = ['proboscis_tip', 'proboscis_head_joint', 'head_torso_joint', 'right_wing_hinge',
+ # 'left_wing_hinge', 'torso_abdomen_joint', 'abdomen_middle', 'abdomen_tip']
+
+ @staticmethod
+ def from_dict(dictionary):
+ module_sets = ModuleSettings()
+ for key, value in dictionary.items():
+ setattr(module_sets, key, value)
+
+ return module_sets
-def init_body_params():
+def init_params():
"""
Returns:
@@ -41,7 +67,7 @@ def init_body_params():
return body_params
-def init_body_bounds():
+def init_bounds():
"""
Returns:
@@ -71,7 +97,7 @@ def init_body_bounds():
return body_bounds
-def init_body_skeleton3d(body_params):
+def init_skeleton3d(body_params):
"""
Returns:
@@ -88,12 +114,12 @@ def init_body_skeleton3d(body_params):
body_skeleton3d['abdomen_middle'] = np.array([-1.5, 0.0, 0.0])
body_skeleton3d['abdomen_tip'] = np.array([-1.5, 0.0, 0.0])
- body_skeleton3d = build_body_skeleton3d(body_skeleton3d, body_params)
+ body_skeleton3d = build_skeleton3d(body_skeleton3d, body_params)
return body_skeleton3d
-def build_body_skeleton3d(init_body_skeleton3d, body_params):
+def build_skeleton3d(init_body_skeleton3d, body_params):
"""
Args:
@@ -153,7 +179,7 @@ def build_body_skeleton3d(init_body_skeleton3d, body_params):
return body_skeleton3d
-def rotate_body_skeleton3d(body_skeleton3d, body_params):
+def rotate_skeleton3d(body_skeleton3d, body_params):
"""
Args:
@@ -176,7 +202,7 @@ def rotate_body_skeleton3d(body_skeleton3d, body_params):
return body_skeleton3d_world
-def translate_body_skeleton3d(body_skeleton3d, body_params):
+def translate_skeleton3d(body_skeleton3d, body_params):
"""
Args:
@@ -198,7 +224,7 @@ def translate_body_skeleton3d(body_skeleton3d, body_params):
return body_skeleton3d_world
-def estimate_body_parameters_from_skeleton3d(body_skeleton3d_world, init_body_params):
+def estimate_params_from_skeleton3d(body_skeleton3d_world, init_body_params):
"""
Estimate parameters of the transformations/rotations needed to get body_skeleton3d_world
@@ -278,7 +304,7 @@ def estimate_body_parameters_from_skeleton3d(body_skeleton3d_world, init_body_pa
return body_params
-def get_copy_body_params_zero(body_params, label_to_zero=['yaw_a', 'pitch_a', 'roll_a']):
+def get_copy_params_zero(body_params, label_to_zero=['yaw_a', 'pitch_a', 'roll_a']):
"""
Args:
@@ -296,7 +322,7 @@ def get_copy_body_params_zero(body_params, label_to_zero=['yaw_a', 'pitch_a', 'r
return new_body_params
-def plot_body(body_skeleton3d):
+def plot(body_skeleton3d):
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
for label in body_skeleton3d.keys():
@@ -309,7 +335,7 @@ def plot_body(body_skeleton3d):
plt.show()
-def residuals_body2d(body_skeleton3d, body_skeleton2d, default_residual, dlt_coefs):
+def residuals2d(body_skeleton3d, body_skeleton2d, default_residual, dlt_coefs):
"""
Compute residuals when comparing body_skeleton3d and body_skeleton2d
@@ -339,7 +365,7 @@ def residuals_body2d(body_skeleton3d, body_skeleton2d, default_residual, dlt_coe
return residuals.flatten()
-def residuals_body3d(body_skeleton3d_1, body_skeleton3d_2, default_residual):
+def residuals3d(body_skeleton3d_1, body_skeleton3d_2, default_residual):
"""
Compute residuals when comparing body_skeleton3d_1 and body_skeleton3d_2
diff --git a/skeleton_fitter/skeleton_modules/insect_hybrid_wings_body_slim.py b/skeleton_fitter/skeleton_modules/insect_hybrid_wings_body_slim.py
index 98067e1698a40d461d313bd86eaf5b56d939aa27..62559f8d66029c9ffb17b69ee634f01d846a99f9 100644
--- a/skeleton_fitter/skeleton_modules/insect_hybrid_wings_body_slim.py
+++ b/skeleton_fitter/skeleton_modules/insect_hybrid_wings_body_slim.py
@@ -1,45 +1,63 @@
import numpy as np
+from dataclasses import dataclass
+
import matplotlib.pyplot as plt
from skeleton_fitter.utils import set_axes3d_equal
-from skeleton_fitter.skeleton_modules.insect_body_slim import init_body_params, init_body_bounds, init_body_skeleton3d, \
- build_body_skeleton3d
+import skeleton_fitter.skeleton_modules.insect_body_slim as body
+
+
+@dataclass
+class ModuleSettings:
+ threshold_likelihood = 0.9
+ threshold_nb_pts = 4
+
+ hybrid_params_init = hybrid.init_params()
+ hybrid_skeleton3d_init = hybrid.init_skeleton3d(hybrid_params_init)
+
+ @staticmethod
+ def from_dict(dictionary):
+ module_sets = ModuleSettings()
+ for key, value in dictionary.items():
+ setattr(module_sets, key, value)
+
+ return module_sets
-def init_hybrid_params():
+def init_params():
"""
Returns:
hybrid_params:
"""
- hybrid_params = init_body_params()
+ hybrid_params = body.init_params()
hybrid_params['wbaverage_span'] = 0.0025
hybrid_params['wbaverage_deviation_a'] = 0.0
return hybrid_params
-def init_hybrid_bounds():
+def init_bounds():
"""
Returns:
hybrid_bounds:
"""
- hybrid_bounds = init_body_bounds()
+ hybrid_bounds = body.init_bounds()
hybrid_bounds['wbaverage_span'] = (0.0005, 0.01)
hybrid_bounds['wbaverage_deviation_a'] = (-45.0, 45.0)
return hybrid_bounds
-def init_hybrid_skeleton3d(hybrid_params):
+def init_skeleton3d(hybrid_params):
"""
Returns:
hybrid_skeleton3d:
"""
- hybrid_skeleton3d = init_body_skeleton3d(hybrid_params)
+ hybrid_skeleton3d = body.init_skeleton3d(hybrid_params)
hybrid_skeleton3d['right_wing_tip'] = \
hybrid_skeleton3d['right_wing_hinge'] \
+ np.array([0.0, - hybrid_params['wbaverage_span'] * np.cos(np.deg2rad(hybrid_params['wbaverage_deviation_a'])),
@@ -57,7 +75,7 @@ def init_hybrid_skeleton3d(hybrid_params):
return hybrid_skeleton3d
-def build_hybrid_skeleton3d(init_hybrid_skeleton3d, hybrid_params):
+def build_skeleton3d(init_hybrid_skeleton3d, hybrid_params):
"""
Args:
@@ -68,7 +86,7 @@ def build_hybrid_skeleton3d(init_hybrid_skeleton3d, hybrid_params):
hybrid_skeleton3d:
"""
- hybrid_skeleton3d = build_body_skeleton3d(init_hybrid_skeleton3d, hybrid_params)
+ hybrid_skeleton3d = body.build_skeleton3d(init_hybrid_skeleton3d, hybrid_params)
hybrid_skeleton3d['right_wing_tip'] = \
hybrid_skeleton3d['right_wing_hinge'] \
+ np.array([0.0, - hybrid_params['wbaverage_span'] * np.cos(np.deg2rad(hybrid_params['wbaverage_deviation_a'])),
@@ -82,7 +100,7 @@ def build_hybrid_skeleton3d(init_hybrid_skeleton3d, hybrid_params):
return hybrid_skeleton3d
-def plot_body_and_wings(body_skeleton3d, wings_skeleton3d):
+def plot(body_skeleton3d, wings_skeleton3d):
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
for label in body_skeleton3d.keys():
diff --git a/skeleton_fitter/skeleton_modules/insect_wings.py b/skeleton_fitter/skeleton_modules/insect_wings.py
index 69e585889f641fa37c98c09a39c4488c9330dad7..e2b4a10f630caeb2719d6b1bd603abefa9d20e0f 100644
--- a/skeleton_fitter/skeleton_modules/insect_wings.py
+++ b/skeleton_fitter/skeleton_modules/insect_wings.py
@@ -2,6 +2,8 @@ import copy
import pandas as pd
import numpy as np
+from dataclasses import dataclass
+
import matplotlib.pyplot as plt
from scipy.spatial.transform import Rotation as R
@@ -11,7 +13,40 @@ from skeleton_fitter.utils import reproject_skeleton3d_to2d, reproject_geometry3
length_from_vector, set_axes3d_equal
-def init_wings_params(body_params):
+@dataclass
+class ModuleSettings:
+ threshold_likelihood = 0.9
+ threshold_nb_pts = 4
+
+ keep_init_aspect_ratio = True
+
+ params_init = wings.init_params(body_params_init)
+ bounds_init = wings.init_bounds(body_bounds_init)
+ skeleton3d_init = wings.init_skeleton3d(params_init)
+
+ body_params_init = body.init_params()
+ body_bounds_init = body.init_bounds()
+ body_skeleton3d_init = body.init_skeleton3d(body_params_init)
+
+ labels = list(skeleton3d_init.keys())
+ # labels = ['hinge', 'leading_edge_q1', 'leading_edge_q2', 'leading_edge_q3', 'tip', 'trailing_edge_q3',
+ # 'trailing_edge_q2', 'trailing_edge_q1']
+ body_labels = list(body_skeleton3d_init.keys())
+
+ labels_to_keep_cst = ['span', 'chord', 'aspect_ratio']
+ body_labels_to_keep_cst = ['proboscis_l', 'torso_l', 'torso_r', 'abdomen_l',
+ 'proboscis_torso_a', 'torso_abdomen_a', 'ratio_com_torso', 'ratio_proboscis_head']
+
+ @staticmethod
+ def from_dict(dictionary):
+ module_sets = ModuleSettings()
+ for key, value in dictionary.items():
+ setattr(module_sets, key, value)
+
+ return module_sets
+
+
+def init_params(body_params):
"""
Args:
@@ -43,7 +78,7 @@ def init_wings_params(body_params):
return wings_params
-def init_wing_bounds(body_bounds):
+def init_bounds(body_bounds):
"""
Args:
@@ -121,8 +156,8 @@ def init_wings_skeleton_geometry3d(wings_params):
len(angle) / len(wings_skeleton3d[side].keys()))
wings_geometry3d[side]['index_prev_labels'] = np.sort(wings_geometry3d[side]['index_prev_labels'])
- wings_skeleton3d[side] = build_wing_skeleton3d(wings_geometry3d[side], wings_params[side])
- wings_geometry3d[side] = build_wing_geometry3d(wings_geometry3d[side], wings_params[side])
+ wings_skeleton3d[side] = build_skeleton3d(wings_geometry3d[side], wings_params[side])
+ wings_geometry3d[side] = build_geometry3d(wings_geometry3d[side], wings_params[side])
return wings_skeleton3d, wings_geometry3d
@@ -201,13 +236,13 @@ def init_wings_skeleton_geometry3d_from_csv(wings_params, csv_path):
'index_prev_labels': index_prev_labels}}
for side in wings_geometry3d.keys():
- wings_skeleton3d[side] = build_wing_skeleton3d(wings_skeleton3d[side], wings_params[side])
- wings_geometry3d[side] = build_wing_geometry3d(wings_geometry3d[side], wings_params[side])
+ wings_skeleton3d[side] = build_skeleton3d(wings_skeleton3d[side], wings_params[side])
+ wings_geometry3d[side] = build_geometry3d(wings_geometry3d[side], wings_params[side])
return wings_skeleton3d, wings_geometry3d
-def build_wing_skeleton3d(init_wing_skeleton3d, wing_params):
+def build_skeleton3d(init_wing_skeleton3d, wing_params):
"""
Args:
@@ -239,7 +274,7 @@ def build_wing_skeleton3d(init_wing_skeleton3d, wing_params):
return wing_skeleton3d
-def build_wing_geometry3d(wing_geometry3d, wing_params):
+def build_geometry3d(wing_geometry3d, wing_params):
"""
Args:
@@ -272,7 +307,7 @@ def build_wing_geometry3d(wing_geometry3d, wing_params):
return wing_geometry3d
-def rotate_and_translate_wing_skeleton3d(wing_skeleton3d, wing_params, side):
+def rotate_and_translate_skeleton3d(wing_skeleton3d, wing_params, side):
"""
Args:
@@ -317,7 +352,7 @@ def rotate_and_translate_wing_skeleton3d(wing_skeleton3d, wing_params, side):
return wing_skeleton3d_world
-def translate_wing_skeleton3d(wing_skeleton3d, wing_params):
+def translate_skeleton3d(wing_skeleton3d, wing_params):
"""
Args:
@@ -340,7 +375,7 @@ def translate_wing_skeleton3d(wing_skeleton3d, wing_params):
return wing_skeleton3d_world
-def rotate_and_translate_wing_geometry3d(wing_geometry3d, wing_params, side):
+def rotate_and_translate_geometry3d(wing_geometry3d, wing_params, side):
"""
Args:
@@ -380,7 +415,7 @@ def rotate_and_translate_wing_geometry3d(wing_geometry3d, wing_params, side):
return wing_geometry3d_world
-def translate_wing_geometry3d(wing_geometry3d, wing_params):
+def translate_geometry3d(wing_geometry3d, wing_params):
"""
Args:
@@ -401,7 +436,7 @@ def translate_wing_geometry3d(wing_geometry3d, wing_params):
return wing_geometry3d_world
-def estimate_wing_parameters_from_skeleton3d(wing_skeleton3d_world, body_params, init_wing_params, side):
+def estimate_params_from_skeleton3d(wing_skeleton3d_world, body_params, init_wing_params, side):
"""
Estimate parameters of the transformations/rotations needed to get wing_skeleton3d_world
@@ -479,7 +514,7 @@ def estimate_wing_parameters_from_skeleton3d(wing_skeleton3d_world, body_params,
return wing_params
-def residuals_wing2d(wing_skeleton3d, wing_skeleton2d, default_residual, dlt_coefs):
+def residuals2d(wing_skeleton3d, wing_skeleton2d, default_residual, dlt_coefs):
"""
Compute residuals when comparing wing_skeleton3d to wing_skeleton2d
@@ -508,7 +543,7 @@ def residuals_wing2d(wing_skeleton3d, wing_skeleton2d, default_residual, dlt_coe
return residuals.flatten()
-def residuals_wing3d(wing_skeleton3d_1, wing_skeleton3d_2, default_residual):
+def residuals3d(wing_skeleton3d_1, wing_skeleton3d_2, default_residual):
"""
Compute residuals when comparing wing_skeleton3d_1 to wing_skeleton3d_2
@@ -530,7 +565,7 @@ def residuals_wing3d(wing_skeleton3d_1, wing_skeleton3d_2, default_residual):
return residuals
-def residuals_wing2d_geometry3d(wing_geometry3d, wing_skeleton2d, default_residual, dlt_coefs):
+def residuals2d_geometry3d(wing_geometry3d, wing_skeleton2d, default_residual, dlt_coefs):
"""
Compute residuals when comparing wing_geometry3d to wing_skeleton2d
@@ -578,7 +613,7 @@ def residuals_wing2d_geometry3d(wing_geometry3d, wing_skeleton2d, default_residu
return residuals.flatten()
-def residuals_wing3d_geometry3d(wing_geometry3d, wing_skeleton3d, default_residual):
+def residuals3d_geometry3d(wing_geometry3d, wing_skeleton3d, default_residual):
"""
Compute residuals when comparing wing_geometry3d to wing_skeleton3d
@@ -619,7 +654,7 @@ def residuals_wing3d_geometry3d(wing_geometry3d, wing_skeleton3d, default_residu
return residuals
-def plot_wing(wing_skeleton3d):
+def plot(wing_skeleton3d):
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
for label in wing_skeleton3d.keys():
@@ -633,7 +668,7 @@ def plot_wing(wing_skeleton3d):
plt.show()
-def plot_wings(wings_skeleton3d):
+def plot_multi(wings_skeleton3d):
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
for side in ['right', 'left']:
@@ -648,7 +683,7 @@ def plot_wings(wings_skeleton3d):
plt.show()
-def plot_wings_geometry3d(wings_geometry3d):
+def plot_multi_geometry3d(wings_geometry3d):
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
for side in ['right', 'left']:
@@ -662,7 +697,7 @@ def plot_wings_geometry3d(wings_geometry3d):
plt.show()
-def get_copy_wings_params_zero(wings_params, label_to_zero=['stroke_a', 'deviation_a', 'rotation_a']):
+def get_copy_multi_params_zero(wings_params, label_to_zero=['stroke_a', 'deviation_a', 'rotation_a']):
"""
Args:
diff --git a/skeleton_fitter/utils.py b/skeleton_fitter/utils.py
index 1c5c3cbb5506ef32b5f36afa3d5e7356d50ce26f..acba887cf510773eadc62c678c29b66a88a8585a 100644
--- a/skeleton_fitter/utils.py
+++ b/skeleton_fitter/utils.py
@@ -1,6 +1,6 @@
import numpy as np
-from cam import calib
+from camera import calib
def reproject_skeleton3d_to2d(skeleton3d, dlt_coefs):
@@ -13,6 +13,7 @@ def reproject_skeleton3d_to2d(skeleton3d, dlt_coefs):
Returns:
skeleton2d_from3d:
"""
+
nb_cam = len(dlt_coefs)
skeleton2d_from3d = {}
diff --git a/tests/processes_preprocessing.yaml b/tests/processes_preprocessing.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..78e0e8b3bfa2a43c9d2c7ef5bef9319d2fa25c96
--- /dev/null
+++ b/tests/processes_preprocessing.yaml
@@ -0,0 +1,67 @@
+0:
+ fn: initialization
+ image_sizes:
+ 1: [256, 256]
+ 2: [256, 256]
+ 3: [256, 256]
+ last_frame: 5451
+ missing_frames:
+ 1: []
+ 2: []
+ 3: []
+ nb_cam: 3
+ paths: {1: ../data/examples/mosquito_escapes/cam1, 2: ../data/examples/mosquito_escapes/cam2,
+ 3: ../data/examples/mosquito_escapes/cam3}
+ process_path: ''
+1:
+ fn: sample
+ kwargs:
+ camn_to_process: [1, 2, 3]
+ delete_previous: true
+ step_frame: 50
+2:
+ fn: track2d
+ kwargs:
+ back_subtractor_settings: {dist2Threshold: 50, shadow_threshold: 1, type: KNN,
+ varThreshold: 12}
+ blob_detector_settings: {area_max: 1000, area_min: 10, aspect_ratio_min: 0.8,
+ eucli_dist_max: 5.0, inertia_ratio_max: 1.0, inertia_ratio_min: 0.0, threshold_max: 255,
+ threshold_min: 1}
+ camn_to_process: [1, 2, 3]
+ from_fn_name: sample
+3:
+ fn: recon3d
+ kwargs:
+ camn_to_process: [1, 2, 3]
+ dlt_path: ../data/calib/examples/mosquito_escapes/20200215_DLTcoefs-py.csv
+ from_fn_name: sample
+ reconstructor_settings: {image_ratio: 4.0, threshold_delay_tracks: 10, threshold_distance_tracks: 0.02,
+ threshold_length_tracks: 5, threshold_mean_repro_err: 5.0}
+4:
+ fn: crop
+ kwargs:
+ camn_to_process: [1, 2, 3]
+ delete_previous: true
+ from_fn_name: sample
+ height_crop: 50
+ width_crop: 50
+5:
+ fn: rotate
+ kwargs:
+ camn_to_process: [2]
+ degrees: 270
+ from_fn_name: crop
+6:
+ fn: stitch
+ kwargs:
+ camn_to_process: [1, 2, 3]
+ delete_previous: true
+ from_fn_name: crop
+7:
+ fn: save_avi
+ kwargs:
+ camn_to_process: [1, 2, 3]
+ delete_previous: true
+ frame_rate: 24
+ from_fn_name: stitch
+ lossless: false
diff --git a/tests/test_calib.py b/tests/test_calib.py
index d75a2a2164cd88413573176c13407bb8e4c194db..321ad04778d8de72eb0ba16eb43bea951d1e017a 100644
--- a/tests/test_calib.py
+++ b/tests/test_calib.py
@@ -1,6 +1,7 @@
+import os
import numpy as np
-from cam import calib
+from camera import calib
def test_dlt2d_1views():
@@ -15,18 +16,19 @@ def test_dlt2d_1views():
uv1 = [[1302, 1147], [1110, 976], [1411, 863], [1618, 1012]]
# Use only one view to perform a 2D calibration of the camera with 4 points of the square:')
- nd = 2
- nc = 1
+ nb_dim = 2
+ nb_cam = 1
# Camera calibration parameters based on view #1 and error of the calibration of view #1 (in pixels):')
- L1, err1 = calib.calib(nd, xy, uv1)
+ dlt_coef1, repro_err1 = calib.calib(nb_dim, xy, uv1)
- assert err1 <= max_mean_err_px
+ assert len(dlt_coef1) == 8
+ assert repro_err1 <= max_mean_err_px
# Reconstruction of the same 4 points based on one view and the camera calibration parameters:
xy1 = np.zeros((len(xy), 2))
for i in range(len(uv1)):
- xy1[i, :] = calib.recon3D(nd, nc, L1, uv1[i])
+ xy1[i, :] = calib.recon3d(nb_dim, nb_cam, dlt_coef1, uv1[i])
# Mean error of the point reconstruction using the DLT (error in cm):
mean_err = np.mean(np.sqrt(np.sum((np.array(xy1) - np.array(xy)) ** 2, 1)))
@@ -49,21 +51,22 @@ def test_dlt2d_2views():
uv2 = [[1094, 1187], [1130, 956], [1514, 968], [1532, 1187]]
# Use 2 views to perform a 2D calibration of the camera with 4 points of the square:
- nd = 2
- nc = 2
+ nb_dim = 2
+ nb_cam = 2
# Camera calibration parameters based on view #1 and error of the calibration of view #1 (in pixels)
- L1, err1 = calib.calib(nd, xy, uv1)
- L2, err2 = calib.calib(nd, xy, uv2)
+ dlt_coef1, repro_err1 = calib.calib(nb_dim, xy, uv1)
+ dlt_coef2, repro_err2 = calib.calib(nb_dim, xy, uv2)
- assert err1 <= max_mean_err_px
- assert err2 <= max_mean_err_px
+ assert len(dlt_coef1) == 8
+ assert repro_err1 <= max_mean_err_px
+ assert repro_err2 <= max_mean_err_px
# Reconstruction of the same 4 points based on 2 views and the camera calibration parameters:
xy12 = np.zeros((len(xy), 2))
- L12 = [L1, L2]
+ dlt_coefs = [dlt_coef1, dlt_coef2]
for i in range(len(uv1)):
- xy12[i, :] = calib.recon3D(nd, nc, L12, [uv1[i], uv2[i]])
+ xy12[i, :] = calib.recon3d(nb_dim, nb_cam, dlt_coefs, [uv1[i], uv2[i]])
# Mean error of the point reconstruction using the DLT (error in cm):
mean_err = np.mean(np.sqrt(np.sum((np.array(xy12) - np.array(xy)) ** 2, 1)))
@@ -90,25 +93,26 @@ def test_dlt3d():
[1661, 1520]]
# Use 4 views to perform a 3D calibration of the camera with 8 points of the cube:
- nd = 3
- nc = 4
+ nb_dim = 3
+ nb_cam = 4
# Camera calibration parameters based on view #1 and error of the calibration of view #1 (in pixels)
- L1, err1 = calib.calib(nd, xyz, uv1)
- L2, err2 = calib.calib(nd, xyz, uv2)
- L3, err3 = calib.calib(nd, xyz, uv3)
- L4, err4 = calib.calib(nd, xyz, uv4)
+ dlt_coef1, repro_err1 = calib.calib(nb_dim, xyz, uv1)
+ dlt_coef2, repro_err2 = calib.calib(nb_dim, xyz, uv2)
+ dlt_coef3, repro_err3 = calib.calib(nb_dim, xyz, uv3)
+ dlt_coef4, repro_err4 = calib.calib(nb_dim, xyz, uv4)
- assert err1 <= max_mean_err_px
- assert err2 <= max_mean_err_px
- assert err3 <= max_mean_err_px
- assert err4 <= max_mean_err_px
+ assert len(dlt_coef1) == 11
+ assert repro_err1 <= max_mean_err_px
+ assert repro_err2 <= max_mean_err_px
+ assert repro_err3 <= max_mean_err_px
+ assert repro_err4 <= max_mean_err_px
# Reconstruction of the same 8 points based on 4 views and the camera calibration parameters:
xyz1234 = np.zeros((len(xyz), 3))
- L1234 = [L1, L2, L3, L4]
+ L1234 = [dlt_coef1, dlt_coef2, dlt_coef3, dlt_coef4]
for i in range(len(uv1)):
- xyz1234[i, :] = calib.recon3D(nd, nc, L1234, [uv1[i], uv2[i], uv3[i], uv4[i]])
+ xyz1234[i, :] = calib.recon3d(nb_dim, nb_cam, L1234, [uv1[i], uv2[i], uv3[i], uv4[i]])
# Mean error of the point reconstruction using the DLT (error in cm):
mean_err = np.mean(np.sqrt(np.sum((np.array(xyz1234) - np.array(xyz)) ** 2, 1)))
@@ -118,19 +122,46 @@ def test_dlt3d():
'is larger than {0} ({1})'.format(max_mean_err_cm, mean_err)
# 2D reprojection of the 8 points
- uv1_repro = calib.find2d(1, L1, xyz1234)
- uv2_repro = calib.find2d(1, L2, xyz1234)
- uv3_repro = calib.find2d(1, L3, xyz1234)
- uv4_repro = calib.find2d(1, L4, xyz1234)
+ uv1_repro = calib.find2d(1, dlt_coef1, xyz1234)
+ uv2_repro = calib.find2d(1, dlt_coef2, xyz1234)
+ uv3_repro = calib.find2d(1, dlt_coef3, xyz1234)
+ uv4_repro = calib.find2d(1, dlt_coef4, xyz1234)
# Mean error of the point reprojection using the DLT (error in px):
- err1_repro = np.sqrt(np.mean(np.sum((uv1 - uv1_repro) ** 2, 1)))
- err2_repro = np.sqrt(np.mean(np.sum((uv2 - uv2_repro) ** 2, 1)))
- err3_repro = np.sqrt(np.mean(np.sum((uv3 - uv3_repro) ** 2, 1)))
- err4_repro = np.sqrt(np.mean(np.sum((uv4 - uv4_repro) ** 2, 1)))
-
- assert err1_repro <= max_mean_err_px
- assert err2_repro <= max_mean_err_px
- assert err3_repro <= max_mean_err_px
- assert err4_repro <= max_mean_err_px
+ repro_err1 = np.sqrt(np.mean(np.sum((uv1 - uv1_repro) ** 2, 1)))
+ repro_err2 = np.sqrt(np.mean(np.sum((uv2 - uv2_repro) ** 2, 1)))
+ repro_err3 = np.sqrt(np.mean(np.sum((uv3 - uv3_repro) ** 2, 1)))
+ repro_err4 = np.sqrt(np.mean(np.sum((uv4 - uv4_repro) ** 2, 1)))
+
+ assert repro_err1 <= max_mean_err_px
+ assert repro_err2 <= max_mean_err_px
+ assert repro_err3 <= max_mean_err_px
+ assert repro_err4 <= max_mean_err_px
+
+
+def test_gen_dlt_coefs_and_save_in_csv() -> None:
+ max_rmse = 10
+
+ img_height = 1024
+ from_matlab = True
+
+ coord_calib_csv_path = '../data/calib/examples/mosquito_escapes/xyz_calibration_device.csv'
+ pts_calib_csv_path = '../data/calib/examples/mosquito_escapes/20200306_xypts.csv'
+
+ dlt_coefs, rmse_errs = calib.gen_dlt_coefs_from_paths(coord_calib_csv_path, pts_calib_csv_path, from_matlab=from_matlab, img_height=img_height)
+
+ assert all(rmse_errs <= max_rmse)
+
+ calib.save_dlt_coefs(dlt_coefs, '../data/calib/examples/mosquito_escapes/20200306_DLTcoefs-py.csv')
+
+
+def main():
+ test_dlt2d_1views
+ test_dlt2d_2views
+ test_dlt3d
+ test_gen_dlt_coefs_and_save_in_csv
+
+
+if __name__ == "__main__":
+ main()
diff --git a/tests/test_images.py b/tests/test_images.py
index 4a195543feac74c6c78d0b882cf5da4e875f69db..a8af6cef1c7fe93fdaa2fc1474881464442d2105 100644
--- a/tests/test_images.py
+++ b/tests/test_images.py
@@ -1,9 +1,132 @@
-from images import process
+import os.path
+import matplotlib.pyplot as plt
+import numpy as np
-def test_process():
- # TODO! Make test_process
- print('TODO')
- # assert "something" is True, "error message"
- # assert tracker2d.test() is True, "error message"
+from images import Image, ImageSequence, MultiImageSequence
+
+def test_image_write() -> None:
+ file_image = "../data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013501.jpg"
+ file_dest = "/cam1_20200216_0759013501_write_test.jpg"
+
+ image = Image(file=file_image)
+
+ image_dup = image.copy()
+ image_dup.write(file_dest)
+
+ image_written = Image(file=file_dest)
+
+ print(image == image_written)
+
+
+# def test_image_file() -> None:
+# file_image = "../data/examples/mosquito_escape/cam1_20200216_075901-selection/cam1_20200216_0759013501.jpg"
+# image = Image(file=file_image, use_caching=True)
+# print(image._data._cache)
+#
+# plt.figure()
+# plt.imshow(image.gradient(return_components=True)[1])
+# plt.show()
+#
+# print(image._data._cache)
+#
+# image.clear_cache()
+#
+# print(image._data._cache)
+
+
+# def test_image_array() -> None:
+# file_image = "../data/examples/mosquito_escape/cam1_20200216_075901-selection/cam1_20200216_0759013501.jpg"
+#
+# shape = (1024, 1024)
+# image = Image(shape=shape)
+# image_file = Image(file=file_image)
+#
+# image_add = image + image_file
+# image_add = image_add - image_file
+# image_add = image_add + 1.2
+#
+# plt.figure()
+# plt.imshow((image_add / 2).get())
+# plt.colorbar()
+# plt.show()
+
+
+def test_image_normalise() -> None:
+ shape = (10, 10)
+ data = np.ones(shape)
+ data[0, 0] = -2.0
+
+ image = Image(data=data)
+ print(image.normalise(minimum=0.0, maximum=1.0).get())
+
+
+def test_image_logical() -> None:
+ test = np.full((10, 10), True, dtype=np.bool_)
+ test[2, 4] = False
+
+ image = Image(data=test, type="bw")
+ print(image.get())
+
+ image.write("test.tif", type_file="tiff_bw")
+
+
+def test_load_image_grey() -> None:
+ file_grey = "../data/examples/mosquito_escapes/cam1/20200216_075901-selection/cam1_20200216_0759013501.jpg"
+
+ image = Image(file=file_grey, type="grey")
+
+ plt.figure()
+ plt.imshow(image.get())
+ plt.show()
+
+
+def test_image_sequence() -> None:
+ seq = ImageSequence.from_directory("../data/examples/mosquito_escapes/cam1/cam1_20200216_075901-sample_binned")
+
+ plt.figure()
+ plt.imshow(seq.median().get())
+ plt.show()
+
+
+def test_multi_image_sequence() -> None:
+ path = '../data/examples/mosquito_escapes/'
+ seq1 = ImageSequence.from_directory(os.path.join(path, 'cam1/cam1_20200216_075901-sample_binned'), use_caching=False)
+ seq2 = ImageSequence.from_directory(os.path.join(path, 'cam2/cam2_20200216_075905-sample_binned'), use_caching=False)
+ seq3 = ImageSequence.from_directory(os.path.join(path, 'cam3/cam3_20200216_075905-sample_binned'), use_caching=False)
+
+ seqs = {'cam1': seq1, 'cam2': seq2, 'cam3': seq3}
+ frame_rates = {'cam1': 12500, 'cam2': 12500, 'cam3': 12500}
+ shifts = {'cam1': 0, 'cam2': 0, 'cam3': 0}
+
+ multi_seq = MultiImageSequence(seqs, frame_rates, shifts)
+
+ assert multi_seq.nb_sequences == 3
+
+ # Or
+ paths = [os.path.join(path, 'cam1/cam1_20200216_075901-sample_binned'),
+ os.path.join(path, 'cam2/cam2_20200216_075905-sample_binned'),
+ os.path.join(path, 'cam3/cam3_20200216_075905-sample_binned')]
+
+ multi_seq = MultiImageSequence.from_directory(paths, 12500)
+
+ assert multi_seq.nb_sequences == 3
+
+ assert multi_seq.get_path() == paths
+ assert multi_seq.names[0] == 'cam1_20200216_075901-sample_binned'
+
+
+def main():
+ # test_image_file()
+ # test_image_array()
+ # test_image_write()
+ # test_image_normalise()
+ # test_image_logical()
+ test_load_image_grey()
+ test_image_sequence()
+ test_multi_image_sequence()
+
+
+if __name__ == "__main__":
+ main()
diff --git a/tests/test_point_tracker.py b/tests/test_point_tracker.py
index 7629978050773466a52271b80e1b07abd42d9b7e..ad1ecc493e4118a9c9efa3ec3ec696f774a4f0e6 100644
--- a/tests/test_point_tracker.py
+++ b/tests/test_point_tracker.py
@@ -1,7 +1,8 @@
-import os, sys
+import os
import numpy as np
-from point_tracker import tracker2d, reconstructor3d
+from point_tracker import Tracker2D, Reconstructor3D
+from camera.calib import read_dlt_coefs
def test_tracker2d():
@@ -9,6 +10,8 @@ def test_tracker2d():
# What to assert?
print('TODO make test_tracker2d that use minimum data')
+ tracker = Tracker2D()
+
# image_format = 'tif'
# nb_cam = 3
#
@@ -16,7 +19,6 @@ def test_tracker2d():
# image_path = os.path.join(os.getcwd(), '../data/examples/mosquito_escape/')
#
# tracker = tracker2d.Tracker2D()
- # tracker.initialize()
#
# for camn in range(1, nb_cam + 1):
# print(os.path.join(image_path, rec_names[camn]))
@@ -31,47 +33,22 @@ def test_tracker2d():
def test_reconstructor3d():
max_rmse = 10
- nb_cam = 3
- img_size = (1024, 1024)
-
- coord_calib_csv = np.genfromtxt(os.path.join(os.getcwd(),
- '../data/calib/examples/mosquito_escape/xyz_calibration_device.csv'),
- delimiter=',', skip_header=0, names=True)
- pts_calib_csv = np.genfromtxt(os.path.join(os.getcwd(),
- '../data/calib/examples/mosquito_escape/20200306_xypts.csv'),
- delimiter=',', skip_header=0, names=True)
-
rec_names = {1: 'cam1_20200303_030117', 2: 'cam2_20200303_030120', 3: 'cam3_20200303_030120'}
- image_path = os.path.join(os.getcwd(), '../data/examples/mosquito_escape/')
+ image_paths = {1: os.path.join(os.getcwd(), '../data/examples/mosquito_escapes/cam1/'),
+ 2: os.path.join(os.getcwd(), '../data/examples/mosquito_escapes/cam2/'),
+ 3: os.path.join(os.getcwd(), '../data/examples/mosquito_escapes/cam3/')}
+ calib_path = '../data/calib/examples/mosquito_escapes/20200306_DLTcoefs-py.csv'
- coord_calib = {'x': np.array(coord_calib_csv['x']), 'y': np.array(coord_calib_csv['y']),
- 'z': np.array(coord_calib_csv['z'])}
-
- reconstructor = reconstructor3d.Reconstructor3D(nb_cam, img_size)
-
- # generate calibration (DLT coefficients)
- pts_calib = {}
- for camn in range(1, nb_cam + 1):
- pts_calib[camn] = {'x': np.array(pts_calib_csv['cam{0}_X'.format(camn)]),
- 'y': np.array(pts_calib_csv['cam{0}_Y'.format(camn)])}
+ dlt_coefs = read_dlt_coefs(calib_path)
- # Fill img_pts_calib with pixel coordinates from Matlab tracking (have to transform coordinate system)
- img_pts_calib = np.array([pts_calib[camn]['x'], img_size[1] - pts_calib[camn]['y']], np.float32).T
+ nb_cam = len(rec_names.keys())
- # Remove np.nan
- index = ~np.isnan(img_pts_calib).any(axis=1)
- img_pts_calib = img_pts_calib[index]
-
- obj_coord_calib = np.array([coord_calib['x'][index], coord_calib['y'][index], coord_calib['z'][index]]).T
-
- _, rmse = reconstructor.gen_dltcoef(camn, obj_coord_calib, img_pts_calib)
-
- assert rmse <= max_rmse, 'mean reprojection error of cam{0} is {1:.2f} pixels (> {2:.2f}'.format(camn, rmse, max_rmse)
+ reconstructor = Reconstructor3D(dlt_coefs)
# From .csv file, fill xs_pose, ys_pose and frames_pose (2d coordinates of blobs)
pts_csv = {}
for camn in range(1, nb_cam + 1):
- pts_csv[camn] = np.genfromtxt(os.path.join(image_path, rec_names[camn], rec_names[camn] + '-2d_points.csv'),
+ pts_csv[camn] = np.genfromtxt(os.path.join(image_paths[camn], rec_names[camn], rec_names[camn] + '-2d_points.csv'),
delimiter=',', skip_header=0, names=True)
xs_coord = np.array([pts_csv[camn]['x_px'] for camn in range(1, nb_cam + 1)])
@@ -81,12 +58,12 @@ def test_reconstructor3d():
xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts, _, = reconstructor.recon_objs(xs_coord, ys_coord, frames_coord)
tracks_dict = reconstructor.recon_tracks(xs_pose, ys_pose, zs_pose, frames_pose, indexes_pts)
- for num_obj in tracks_dict['obj'].keys():
+ for nb_obj in tracks_dict['obj'].keys():
- xs_coord_repro, ys_coord_repro = reconstructor.get_2d_reprojection(tracks_dict['obj'][num_obj]['x'],
- tracks_dict['obj'][num_obj]['y'],
- tracks_dict['obj'][num_obj]['z'])
- index = np.array(tracks_dict['obj'][num_obj]['index_pts']).T
+ xs_coord_repro, ys_coord_repro = reconstructor.get_2d_reprojection(tracks_dict['obj'][nb_obj]['x'],
+ tracks_dict['obj'][nb_obj]['y'],
+ tracks_dict['obj'][nb_obj]['z'])
+ index = np.array(tracks_dict['obj'][nb_obj]['index_pts']).T
for i in range(0, nb_cam):
@@ -100,5 +77,12 @@ def test_reconstructor3d():
assert rmse_y_repro <= max_rmse, 'Mean reprojection y error of tracks is {0:.2f} pixels (> {1:.2f}'.format(rmse_x_repro, max_rmse)
+def main():
+ test_tracker2d
+ test_reconstructor3d
+
+
+if __name__ == "__main__":
+ main()
diff --git a/tests/test_process_batch.py b/tests/test_process_batch.py
new file mode 100644
index 0000000000000000000000000000000000000000..bc3ce53a86d96e7985b203b3b1617b2c83449703
--- /dev/null
+++ b/tests/test_process_batch.py
@@ -0,0 +1,139 @@
+import os
+
+from process.batch_processing import BatchProcessing
+from point_tracker.tracker2d import BlobDetectorSettings, BackgroundSubtractorSettings
+from point_tracker.reconstructor3d import Reconstructor3DSettings
+
+
+def test_batch_processing_images_preprocessing() -> None:
+
+ recording_paths = ['../data/examples/mosquito_escapes/cam1',
+ '../data/examples/mosquito_escapes/cam2',
+ '../data/examples/mosquito_escapes/cam3']
+ save_path = '_Process'
+ dlt_path = os.path.join(os.getcwd(), '../data/calib/examples/mosquito_escapes/20200215_DLTcoefs-py.csv')
+
+ rec_names_to_process = ['20200216_075901-selection'] # names of the recordings to process
+ # (Should be the name of the folder with the images of the main camera (by default: #1))
+
+ framerate = 12500
+
+ img_process = BatchProcessing.from_directory_by_names(recording_paths, framerate, save_path,
+ type_image="grey", threshold_nb_diff_char=2)
+ # img_process = BatchProcessing.from_directory_by_dates(recording_paths, framerate, save_path, type_image="grey",
+ # expr=r"\d{8}_\d{6}", format_date_str='%Y%m%d_%H%M%S', threshold_s=90)
+
+ # -----------------------------------------------------------------------------------------------------------------------
+ # Will prepare recordings to be used by Deeplabcut (do each step separately)
+ # Convert images to 8 bits and reduce frame_rate (save in cam_save_paths/_Sample)
+ img_process.sample_images(50, rec_names_to_process=rec_names_to_process, delete_previous=True)
+
+ # Do 2d tracking (blob detection) on images in cam_save_paths/_Sample
+ blob_detec_sets = BlobDetectorSettings(area_min=10, area_max=1000)
+ img_process.track2d(blob_detector_settings=blob_detec_sets,
+ rec_names_to_process=rec_names_to_process, from_fn_name='sample')
+
+ # Do 3d reconstruction of tracks
+ reconstructor_sets = Reconstructor3DSettings(image_ratio=4.0)
+ # use image_ratio=4.0 because the images to analysed have been binned compared to the images used for the calibration)
+ img_process.recon3d(dlt_path, reconstructor_settings=reconstructor_sets,
+ rec_names_to_process=rec_names_to_process, from_fn_name='sample')
+
+ # Crop all frames and save in cam_save_paths/_Cropped
+ img_process.crop_images('sample', 50, 50, rec_names_to_process=rec_names_to_process, delete_previous=True)
+
+ # Rotate view 2 (to always have piston coming from right side)
+ img_process.rotate_images(270, rec_names_to_process=rec_names_to_process, from_fn_name='crop', camn_to_process=[2])
+
+ # Stitch all views together and save in cam_save_paths/_Stitched
+ img_process.stitch_images(rec_names_to_process=rec_names_to_process, from_fn_name='crop', delete_previous=True)
+
+ # Save each recording in .avi
+ img_process.save_avi(rec_names_to_process=rec_names_to_process, from_fn_name='stitch', delete_previous=True)
+
+
+def test_batch_multiprocessing_images_preprocessing() -> None:
+
+ recording_paths = ['../data/examples/mosquito_escapes/cam1',
+ '../data/examples/mosquito_escapes/cam2',
+ '../data/examples/mosquito_escapes/cam3']
+ save_path = '_Process'
+
+ framerate = 12500
+
+ img_process = BatchProcessing.from_directory_by_names(recording_paths, framerate, save_path,
+ type_image="grey", threshold_nb_diff_char=2)
+
+ rec_names_to_process = ['20200216_075901-selection']
+
+ # -----------------------------------------------------------------------------------------------------------------------
+ # Will prepare recordings to be used by Deeplabcut (do everything automatically by using 'multi_processes')
+ # Do cropping, rotating, stitching, etc at once and save as .avi (save much time and space)
+ img_process.multi_processes('processes_preprocessing.yaml', rec_names_to_process=rec_names_to_process, delete_previous=True)
+
+
+def test_batch_processing_dlc_analysis() -> None:
+ # TODO check test_batch_processing_dlc_analysis with deeplabcut installed
+
+ # -----------------------------------------------------------------------------------------------------------------------
+ # Run Deeplabcut analysis + load tracking result (filtering, unscrambling, and reverse processes) + Reconstruct 3d coord
+ dlc_cfg_path = '/home/user/Desktop/Antoine/_maDLC/config.yaml'
+ dlt_path = os.path.join(os.getcwd(), '../data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs-py.csv')
+
+ rec_names_to_process = ['20200216_075901-selection'] # names of the recordings to process
+
+ shuffle = 31
+ iteration = 70000
+ model_name = 'DLC_resnet50_MatingKinematicsshuffle{0}_{1}'.format(shuffle, iteration)
+
+ img_process = BatchProcessing.from_directory_by_dates()
+
+ # Track features using DeepLabCut
+ img_process.analyse_dlc(dlc_cfg_path, shuffle, 0, 5, model_name, save_avi=False,
+ rec_names_to_process=rec_names_to_process, from_fn_name='save_avi', delete_previous=True)
+
+ # TODO split the following step into several (Loading, Reverse processes and finally 3d reconstruction)??
+ # Load (+ filtering and unscrambling) 2d coords from DLC + Reverse processes (un-stitch, rotate back, un-crop) + Reconstruct 3d coord
+
+ reconstructor_sets = Reconstructor3DSettings(image_ratio=4.0)
+ # use image_ratio=4.0 because the images to analysed have been binned compared to the images used for the calibration)
+ img_process.load_dlc(model_name, 'skeleton', dlt_path, reconstructor_settings=reconstructor_sets,
+ rec_names_to_process=rec_names_to_process, from_fn_name='analyse_dlc')
+
+
+def test_batch_processing_fit_skeleton() -> None:
+ # -----------------------------------------------------------------------------------------------------------------------
+ # Fit 3D skeleton to 2D coordinate from Deeplabcut
+
+ dlt_path = os.path.join(os.getcwd(), '../data/calib/examples/mosquito_escapes/to_delete/20200615_DLTcoefs-py.csv')
+ #csv_wings_geo_path =
+
+ rec_names_to_process = ['20200216_075901-selection']
+
+ shuffle = 31
+ iteration = 70000
+ model_name = 'DLC_resnet50_MatingKinematicsshuffle{0}_{1}'.format(shuffle, iteration)
+
+ res_method = '3d' # '2d', '3d' or '2d_geo'
+ opt_method = 'leastsq' # 'powell' or 'leastsq' or 'least_squares'
+ body_param_names = ['yaw_a', 'pitch_a', 'roll_a', 'x_com', 'y_com', 'z_com']
+ wing_param_names = ['stroke_a', 'deviation_a', 'rotation_a']
+
+ img_process = BatchProcessing.from_directory_by_dates()
+
+ # Optimize fit of skeleton to find body and wings angles
+ img_process.multiprocessing = True
+ img_process.threshold_likelihood = 0.85
+ img_process.fit_skeleton('fly', model_name, res_method, opt_method, dlt_path,
+ rec_names_to_process=rec_names_to_process, from_fn_name='load_dlc')
+
+
+def main():
+ test_batch_processing_images_preprocessing()
+ test_batch_multiprocessing_images_preprocessing()
+ test_batch_processing_dlc_analysis()
+ test_batch_processing_fit_skeleton()
+
+
+if __name__ == "__main__":
+ main()
diff --git a/tests/test_process_utils.py b/tests/test_process_utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..9cccf12db4bfdf406710a341667d8def25c6a2dc
--- /dev/null
+++ b/tests/test_process_utils.py
@@ -0,0 +1,54 @@
+import process.utils as process_utils
+
+
+def test_match_folders() -> None:
+ folder_names = [['cam1_20220304_055123', 'cam1_20220304_101111', 'cam1_20220304_110140', 'cam1_20220304_120352'],
+ ['cam2_20220304_055123', 'cam2_20220304_101115', 'cam2_20220304_120402'],
+ ['cam3_20220304_055123', 'cam3_20220304_101111', 'cam3_20220304_111158', 'cam3_20220304_120352']]
+
+ matched_directories = process_utils.match_folders_by_name(folder_names)
+ assert matched_directories == []
+
+ matched_directories = process_utils.match_folders_by_name(folder_names, threshold_nb_diff_char=1)
+ assert matched_directories == [['cam1_20220304_055123', 'cam2_20220304_055123', 'cam3_20220304_055123']]
+
+ matched_directories = process_utils.match_folders_by_date(folder_names, threshold_s=60)
+ assert matched_directories == [['cam1_20220304_055123', 'cam2_20220304_055123', 'cam3_20220304_055123'],
+ ['cam1_20220304_101111', 'cam2_20220304_101115', 'cam3_20220304_101111'],
+ ['cam1_20220304_120352', 'cam2_20220304_120402', 'cam3_20220304_120352']]
+
+
+def test_match_recordings() -> None:
+ directories = ['../data/examples/mosquito_escapes/cam1/cam1_20200303_030117',
+ '../data/examples/mosquito_escapes/cam2/cam2_20200303_030120',
+ '../data/examples/mosquito_escapes/cam3/cam3_20200303_030120']
+
+ matched_directories = process_utils.match_recordings_by_date(directories, threshold_s=60)
+ assert len(matched_directories) == 3
+
+
+def test_get_unmatched_folder_names() -> None:
+ folder_names = [['cam1_20220304_055123', 'cam1_20220304_101111', 'cam1_20220304_110140', 'cam1_20220304_120352'],
+ ['cam2_20220304_055123', 'cam2_20220304_101115', 'cam2_20220304_120402'],
+ ['cam3_20220304_055123', 'cam3_20220304_101111', 'cam3_20220304_111158', 'cam3_20220304_120352']]
+
+ matched_folder_names = [['cam1_20220304_055123', 'cam2_20220304_055123', 'cam3_20220304_055123'],
+ ['cam1_20220304_101111', 'cam2_20220304_101115', 'cam3_20220304_101111'],
+ ['cam1_20220304_120352', 'cam2_20220304_120402', 'cam3_20220304_120352']]
+
+ unmatched_folder_names = process_utils.get_unmatched_folders(folder_names, matched_folder_names)
+
+ assert unmatched_folder_names == [['cam1_20220304_110140'], [], ['cam3_20220304_111158',]]
+
+
+def main():
+ test_match_folders
+ test_match_recordings
+ test_get_unmatched_folder_names
+
+
+if __name__ == "__main__":
+ main()
+
+
+
diff --git a/tests/test_skeleton_fitter.py b/tests/test_skeleton_fitter.py
index 7cb8fca7cd5d7e4cf6dc272ff0c6aa988dcb560c..0463ced52d091a846f46776b7854d1aa94b6eeab 100644
--- a/tests/test_skeleton_fitter.py
+++ b/tests/test_skeleton_fitter.py
@@ -18,30 +18,30 @@ def test_skeleton_fitter():
dlt_coefs = [[-6.86546837e+02, 1.29528915e+04, 1.02377641e+02, 4.73021647e+02,
-7.28479582e+02, 1.16245489e+02, -1.31706744e+04, 5.05410355e+02,
- -1.75493289e+00, -2.16534662e-01, -8.74025031e-02, 1.00000000e+00],
+ -1.75493289e+00, -2.16534662e-01, -8.74025031e-02],
[ 1.32835992e+04, -1.30629279e+02, -8.78227221e+02, 4.46517571e+02,
-1.21075194e+02, -1.31697555e+04, -8.05482186e+02, 4.24669476e+02,
- 9.01983567e-02, 3.58804475e-02, -1.98002751e+00, 1.00000000e+00],
+ 9.01983567e-02, 3.58804475e-02, -1.98002751e+00],
[-1.28278296e+04, -7.90718730e+02, 9.37964723e+00, 4.30413100e+02,
-6.06304504e+01, -9.31664890e+02, -1.28878959e+04, 5.05777107e+02,
- 1.83401606e-02, -2.16231985e+00, -3.49328202e-02, 1.00000000e+00]]
+ 1.83401606e-02, -2.16231985e+00, -3.49328202e-02]]
res_methods = ['2d', '2d_geo', '3d']
opt_methods = ['nelder-mead', 'powell', 'least_squares', 'leastsq']
sides = ['left', 'right']
- body_params_init = insect_body_slim.init_body_params()
- wings_params_init = insect_wings.init_wings_params(body_params_init)
- hybrid_params_init = insect_hybrid_wings_body_slim.init_hybrid_params()
+ body_params_init = insect_body_slim.init_params()
+ wings_params_init = insect_wings.init_params(body_params_init)
+ hybrid_params_init = insect_hybrid_wings_body_slim.init_params()
- body_bounds = insect_body_slim.init_body_bounds()
- wing_bounds = insect_wings.init_wing_bounds(body_bounds)
- hybrid_bounds_init = insect_hybrid_wings_body_slim.init_body_bounds()
+ body_bounds = insect_body_slim.init_bounds()
+ wing_bounds = insect_wings.init_bounds(body_bounds)
+ hybrid_bounds_init = insect_hybrid_wings_body_slim.init_bounds()
- body_skeleton3d_init = insect_body_slim.init_body_skeleton3d(body_params_init)
+ body_skeleton3d_init = insect_body_slim.init_skeleton3d(body_params_init)
wings_skeleton3d_init, wings_geometry3d_init = insect_wings.init_wings_skeleton_geometry3d_from_csv(wings_params_init, csv_wings_geo_path)
- hybrid_skeleton3d_init = insect_hybrid_wings_body_slim.init_hybrid_skeleton3d(hybrid_params_init)
+ hybrid_skeleton3d_init = insect_hybrid_wings_body_slim.init_skeleton3d(hybrid_params_init)
# Parameters to generate fake 3d and 2d skeletons
body_params = copy.deepcopy(body_params_init)
@@ -57,16 +57,16 @@ def test_skeleton_fitter():
# Test all res_methods and opt_methods
for res_method in res_methods:
- body_skeleton = insect_body_slim.rotate_body_skeleton3d(body_skeleton3d_init, body_params)
- body_skeleton = insect_body_slim.translate_body_skeleton3d(body_skeleton, body_params)
+ body_skeleton = insect_body_slim.rotate_skeleton3d(body_skeleton3d_init, body_params)
+ body_skeleton = insect_body_slim.translate_skeleton3d(body_skeleton, body_params)
- hybrid_skeleton = insect_body_slim.rotate_body_skeleton3d(hybrid_skeleton3d_init, body_params)
- hybrid_skeleton = insect_body_slim.translate_body_skeleton3d(hybrid_skeleton, body_params)
+ hybrid_skeleton = insect_body_slim.rotate_skeleton3d(hybrid_skeleton3d_init, body_params)
+ hybrid_skeleton = insect_body_slim.translate_skeleton3d(hybrid_skeleton, body_params)
wings_skeleton, wings_geometry = copy.deepcopy(wings_skeleton3d_init), copy.deepcopy(wings_geometry3d_init)
for side in sides:
- wings_skeleton[side] = insect_wings.rotate_and_translate_wing_skeleton3d(wings_skeleton3d_init[side], wings_params[side], side)
- wings_geometry[side] = insect_wings.rotate_and_translate_wing_geometry3d(wings_geometry3d_init[side], wings_params[side], side)
+ wings_skeleton[side] = insect_wings.rotate_and_translate_skeleton3d(wings_skeleton3d_init[side], wings_params[side], side)
+ wings_geometry[side] = insect_wings.rotate_and_translate_geometry3d(wings_geometry3d_init[side], wings_params[side], side)
if '2d' in res_method:
body_skeleton = utils.reproject_skeleton3d_to2d(body_skeleton, dlt_coefs)
@@ -123,6 +123,9 @@ def test_skeleton_fitter():
"Too high nb_iterations ({0} > {1}) when optimise fitting in {2} with {3}".format(hybrid_nb_iterations, max_nb_iterations, res_method, opt_method)
-if __name__ == '__main__':
-
+def main():
test_skeleton_fitter()
+
+
+if __name__ == "__main__":
+ main()