A PyQt5 desktop GUI for frame-accurate annotation of rodent behavior videos, forked from Eric Leonardis's original annotation GUI and extended for the TopoMimic research workflow. The fork adds a two-track label schema — five macro behaviors (Immobile, Rear, Turn, Walk, Groom) on a behavior track with an independent Left/Right/Straight direction track above it — and a qpos-driven auto-labeling pass that populates both tracks from per-clip pose CSVs using calibrated rules, so annotators correct a first draft rather than labeling from scratch. Thresholds regenerate from a reference session via a single script, and gesture-based timeline editing with full undo/redo makes fixing the remaining frames fast.
GitHub Repository