Publication code for Fly navigational responses to odor motion and gradient cues are tuned to plume statistics).
The Figure 1 and 3 pipelines deliberately keep large movies and model checkpoints outside Git; their renderers consume compact, provenance-carrying NPZ summaries.
conda env create -f environment.yml
conda activate gradient_motion_multiple_plumes
pip install -e ".[figures123,test]" # adds PyTorch for Figure 3 training/extraction
python scripts/render_published_figure2.py
python -m pytest -qThe behavior source is DANDI dandiset 001871, version 0.260630.1657. Compact published tables are checked in, and scripts/generate_published_summary_tables.py can regenerate them from the archived behavior NWB files.
The DANDI inputs are attributed as: Brudner, Samuel (2026), Data for fly navigational responses exploiting plume-specific odor motion and gradient cues, version 0.260630.1657, DANDI Archive, doi:10.48324/dandi.001871/0.260630.1657, CC BY 4.0. Smooth-plume data come from Dryad doi:10.5061/dryad.g27mq71, CC0.