Files:
{train,validation}.json: Contains annotations for LVIS categories in TAO train and validation.{train,validation}_with_freeform.json: Contains annotations for all categories in TAO train and validation.test_without_annotations.json: JSON for test videos. The 'images' and 'videos' fields contain the images and videos that will be evaluated on the test set.test_categories.json: Contains a list of categories which will be evaluated on the TAO test set.video-lists/: Contains lists of videos and their source datasets contained in TAO. Please do not train on any videos in these lists! Seevideo-lists/README.mdfor details.