Newer
Older
# DDHT AI
Created by @iadriaens
June 23, 2021
## Project description
Knowledge Base Digital Data - High Tech Artificial Intelligence (KB DDHT AI) aims to develop and validate tools for the automated detection and tracking of individual cows and characterization of their behavior based on imagery/video data.
In 2021, the target behaviors are ‘standing up’ and ‘lying down’, from which the duration might be linked to e.g. welfare and claw health. Additionally, a multiple cow-tracker will be implemented that allows to track animals throughout the barn.
This repository will contain the computing pipeline code modules for DDHT AI.
I created a local environment via Anaconda for this project, can be requested via ines.adriaens@wur.nl. The environment configures Python 3.7, and I use Spyder as IDE.
## Codes
__Pre-processing__
1) Frame sampler
2) Frame proprocessing
3) Inventarisation
4) Data selection
5) Data splicing
6) Annotation
7) Data augmentation
__Modelling__
1) Model / training
2) Cross-validation
3) Model evaluation and visualisation
__Post-processing__
1) Tracking (ina)
2) Getting up / lying down