Install_Guide_Adlink_Neon.md



Installation guide of all software on the Adlink NEON smart-camera
Password Adlink (for sudo): adlink 

This installation guide is tested on: Jetpack 4.4 (R32.4.3), CUDA 10.2, Python 3.6, TensorRT 7.1.3.1 

There are several libraries already installed on the Adlink Neon: 

python2, python3, opencv, cuda, numpy 
 

1) Check the versions of the installed packages (in the terminal):

cat /etc/nv_tegra_release (should print the version of Jetpack)

nvcc --version (should the CUDA details) 
 

2) Install Pytorch and Torchvision:


First, check which Pytorch version is compatible with the Jetpack/Cuda version. In this case it's Pytorch version 1.6 (you can Google the compatibility at this URL: https://forums.developer.nvidia.com/t/pytorch-for-jetson-version-1-7-0-now-available/72048) 

Run the following commands in the terminal:

sudo apt install python3-pip
pip3 install wheel
wget https://nvidia.box.com/shared/static/9eptse6jyly1ggt9axbja2yrmj6pbarc.whl -O torch-1.6.0-cp36-cp36m-linux_aarch64.whl
sudo apt-get install python3-pip libopenblas-base libopenmpi-dev
pip3 install Cython
pip3 install numpy torch-1.6.0-cp36-cp36m-linux_aarch64.whl
sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libavcodec-dev libavformat-dev libswscale-dev
git clone --branch v0.7.0 https://github.com/pytorch/vision torchvision
cd torchvision
export BUILD_VERSION=0.7.0
sudo python3 setup.py install (if this doesn't work: python3 setup.py install)
cd ../ 
 

3) Check if Pytorch and Torchvision are correctly installed:

python3
import torch
import torchvision
print(torch.version.cuda) (should print: 10.2)

print('CUDA available: ' + str(torch.cuda.is_available())) (should print: CUDA available: True)

torch.cuda.get_device_name(0) (should print: NVIDIA Tegra X2)

quit() 
 

4) Install the software packages needed for YOLOv5:

pip3 install matplotlib
pip3 install pycocotools
pip3 install tqdm
pip3 install PyYAML==5.3.1
pip3 install pycuda
sudo apt-get install python3-tk 
 

5) Clone the optima/inference code:

git clone https://git.wur.nl/optima/inference.git

cd inference 
 

6) Install the software packages needed for classification (EfficientNet-Pytorch):

wget https://nvidia.box.com/shared/static/49fzcqa1g4oblwxr3ikmuvhuaprqyxb7.whl -O onnxruntime_gpu-1.6.0-cp36-cp36m-linux_aarch64.whl
pip3 install onnxruntime_gpu-1.6.0-cp36-cp36m-linux_aarch64.whl 
 

7) Install the SWIG software (needed for image acquisition):

download swig-3.0.12: https://sourceforge.net/p/swig/news/2017/01/swig-3012-released/

cd swig-3.0.12
./configure --prefix=/home/adlink/library/swigtool
sudo make
sudo make install
sudo gedit /etc/profile
copy-paste these two lines at the end of the file: 

   export SWIG_PATH=/home/adlink/library/swigtool/bin 

   export PATH=SWIG_PATH:PATH 


source /etc/profile
sudo reboot
swig -version (should print: 3.0.12) 
 

8) Install pypylon software (needed for image acquisition):

git clone https://github.com/basler/pypylon.git

cd pypylon
pip3 install .
cd samples
python3 opencv.py 
 

9) Install the GPS software:

pip3 install pyserial
pip3 install pynmea2
sudo usermod -a -G tty $USER
sudo usermod -a -G dialout $USER
sudo reboot 
 

10) OPTIONAL: Install the TensorRT backend for ONNX (to significantly speed up the EfficientNet image classification):

 The description below has been tested for TensorRT version 7.1.3.1. Be aware that this installation requires the presence of the SWIG software (step 7)!

dpkg -l | grep TensorRT (should print a big list with all TensorRT libraries, including versions)

sudo apt-get install python3-pip libprotoc-dev protobuf-compiler
pip3 install onnx==1.6.0
cd inference/onnx-tensorrt
sudo python3 setup.py install 


If there's still an error, like unable to execute 'swig': No such file or directory, then you have to have to run the following command:

swig -python -c++ -modern -builtin -o nv_onnx_parser_bindings_wrap.cpp nv_onnx_parser_bindings.i (should execute without any feedbacks and errors)
sudo python3 setup_2.py install 
 

11) OPTIONAL: Make desktop shortcuts:

follow the instructions in this URL: https://gist.github.com/nathakits/7efb09812902b533999bda6793c5e872

chmod u+x /home/adlink/inference/inference.py
Right click on the desktop-file and select Properties. In the Permissions tab, set the execute to allow executing file as program

[Desktop Entry]

Name=detection

Version=1

Icon=/home/adlink/inference/data/optima.jpg

Exec=python3 /home/adlink/inference/inference.py --cfg /home/adlink/inference/cfg/detection.yaml

Terminal=true

Type=Application

[Desktop Entry]

Name=classification

Version=1

Icon=/home/adlink/inference/data/optima.jpg

Exec=python3 /home/adlink/inference/inference.py --cfg /home/adlink/inference/cfg/classification.yaml

Terminal=true

Type=Application

[Desktop Entry]

Name=classification_patches

Version=1

Icon=/home/adlink/inference/data/optima.jpg

Exec=python3 /home/adlink/inference/inference.py --cfg /home/adlink/inference/cfg/classification_patches.yaml

Terminal=true

Type=Application

[Desktop Entry]

Name=keyboard

Version=1

Icon=/home/adlink/inference/data/keyboard.jpg

Exec=onboard

Terminal=false

Type=Application