Behavior classification#

Annolid supports two complementary behavior workflows:

  1. Event scoring in the GUI (manual or semi-manual), with exports and summaries.

  2. Model-based classification (optional / experimental), for projects that have labeled training data.

2) Train a behavior classifier (optional / experimental)#

Annolid includes behavior training utilities under annolid/behavior/.

The training entry point is:

python -m annolid.behavior.training.train --video_folder /path/to/labeled_clips

Note

The training dataset loader expects per-video CSV annotations (same stem as the video file) with columns such as Behavior and Trial time. If your data is in a different format, you’ll likely want to adapt annolid/behavior/data_loading/datasets.py to your project.