yolov5_tracking / val_utils /docs /YouTube-VIS-format.txt
xfys's picture
Upload 645 files
47af768
raw
history blame
1.5 kB
Taken from: https://competitions.codalab.org/competitions/20128#participate-get-data
The label file follows MSCOCO's style in json format. We adapt the entry name and label format for video. The definition of json file is:
{
"info" : info,
"videos" : [video],
"annotations" : [annotation],
"categories" : [category],
}
video{
"id" : int,
"width" : int,
"height" : int,
"length" : int,
"file_names" : [file_name],
}
annotation{
"id" : int,
"video_id" : int,
"category_id" : int,
"segmentations" : [RLE or [polygon] or None],
"areas" : [float or None],
"bboxes" : [[x,y,width,height] or None],
"iscrowd" : 0 or 1,
}
category{
"id" : int,
"name" : str,
"supercategory" : str,
}
The submission file is also in json format. The file should contain a list of predictions:
prediction{
"video_id" : int,
"category_id" : int,
"segmentations" : [RLE or [polygon] or None],
"score" : float,
}
The submission file should be named as "results.json", and compressed without any subfolder. There is an example "valid_submission_sample.zip" in download links above. The example is generated by our proposed MaskTrack R-CNN algorithm.