Multi-Camera Synchronization and Instance Tracking for Avian Behavior Analysis in a Floating Aviary

Installation Guide

This guide explains how to set up the environment, download required checkpoints, and launch the app

Prerequisites

• Conda (Anaconda or Miniconda)
• Git (optional, if you clone this repo)
• Python 3.10 (the Conda env will install this)
• CUDA-enabled GPU and the correct drivers for faster inference
• nvidia-cuda-toolkit installed (optional, recommended)

1) Clone repository

git clone https://github.com/gergelydinya/AviTrack.git
cd AviTrack

2) Create and activate the Conda environment

# From the project root
conda create -n avitrack python=3.10 -y
conda activate avitrack
pip install -r requirements.txt
# Match to your CUDA version 12.8 ~ cu128
pip install --index-url https://download.pytorch.org/whl/cu128 torch torchvision torchaudio

3) Install GPU-enabled decord

Read https://github.com/dmlc/decord for details

4) Download checkpoints

...

4) Run the software

streamlit run home.py

Citation

If you use AviTrack, please cite our paper:

@article{avitrack,
  title   = {Multi-camera synchronization and instance tracking for Avian behavior analysis in a floating aviary},
  author  = {Dinya, Gergely and Gelencs{\'e}r-Horv{\'a}th, Anna and Ferretti, Andrea and Rattenborg, Niels and L{\H{o}}rincz, Andr{\'a}s},
  journal = {Ecological Informatics},
  volume  = {92},
  pages   = {103419},
  year    = {2025},
  month   = dec,
  doi     = {10.1016/j.ecoinf.2025.103419},
  url     = {https://www.sciencedirect.com/science/article/pii/S1574954125004285}
}