Deep learning from videography as a tool for measuring infection in poultry

Code and data for "Deep learning from videography as a tool for measuring infection in poultry".

Link to the paper: https://doi.org/10.1098/rsos.250151

Link to the data: https://doi.org/10.5281/zenodo.14712491

Setup

Python and R dependencies are managed together via pixi:

pixi install
pixi run post_install   # installs brms and envalysis from CRAN

Activate the environment with pixi shell (or prefix commands with pixi run).

Versions: Python 3.10.11, R 4.4.x.

Python

DeepLabCut data and training

Available soon

Data is available at https://zenodo.org/records/14712492

Feature extraction from DeepLabCut predictions

python -m dlc4ecoli.dlc.extract /path/to/data

Optical flow feature extraction

python -m dlc4ecoli.of.extract /path/to/data

Reproducing the figures

You can reproduce most figures by running the plots.ipynb notebook.

The other brms figures are created from the R script in dlc4ecoli/utils/analysis.R

R

Mixed-effects modelling

Run from the repository root:

Rscript dlc4ecoli/utils/analysis.R

Reference

@article{10.1098/rsos.250151,
  title={Deep learning from videography as a tool for measuring E. coli infection in poultry},
  author={Scheidwasser, Neil and Poulsen, Louise Ladefoged and Leow, Prince Ravi and Khurana, Mark Poulsen and Iglesias-Carrasco, Maider and Laydon, Daniel Joseph and Donnelly, Christl Ann and Bojesen, Anders Miki and Bhatt, Samir and Duch{\^e}ne, David Alejandro},
  journal={Royal Society Open Science},
  volume={12},
  number={10},
  pages={250151},
  year={2025},
  publisher={The Royal Society}
}