Everything stays in your browser. Load a sample video, mark events with the conventions of the task
(start/end per class), export ground_truth.json and score with
python evaluate.py --pred predictions_samples.json --gt ground_truth.json --per-video.
| start | end | label |
|---|