This repository contains tools for the KABR dataset preparation.
detector2cvat.py:
Detect objects with Ultralytics YOLO detections, apply SORT tracking and convert tracks to CVAT format.
python detector2cvat.py path_to_videos path_to_save
cvat2ultralytics.py:
Convert CVAT annotations to Ultralytics YOLO dataset.
python cvat2ultralytics.py path_to_videos path_to_annotations dataset_name [skip_frames]
tracks_extractor.py:
Extract mini-scenes from CVAT tracks.
python tracks_extractor.py path_to_videos path_to_annotations [tracking]
player.py:
Player for track and behavior observation.
python player.py path_to_folder [save]
cvat2slowfast.py:
Convert CVAT annotations to the dataset in Charades format.
python cvat2slowfast.py path_to_mini_scenes dataset_name [zebra, giraffe]