The video is just a pedestrian standing still at the edge of the road.
Image stream of test: https://drive.google.com/drive/folders/192i4Sq1M8D8pOV8Rss9YLppZZ4DJnjQL?usp=sharing
where [0…4] refer to different tracks, and AAATotal refers to the number of classifications made for that track, and the rest of the keys are the proportion of each class.
Total is the sum accross all tracks.
Break is the greatest time break between classifications.