The bounding box annotations for A2D are in the json file a2d_bounding_box_annotations.json. They are a dictionary in the following format: { '$video_name$': [{ 'sentence': _, 'instance': _, 'boxes': { '$frame$': (x, y, w, h) } }] } $video_name$ is the name of the video. $frame$ is the index of a given frame in the video, within the range [0, n_frames-1]. The bounding boxes are represented by their top left point (x, y) and their width/height (w, h).