[Dataset] Berkeley Deep Drive to YOLO

Berkeley Deep Drive

https://bdd-data.berkeley.edu/portal.html#download

Images : 100k images
Annotation : Detection 2020 Labels

BDD toolkit (Github)

change BDD annotation file format to COCO dataset format

#toolkit
https://github.com/bdd100k/bdd100k
#label convert docs
https://doc.bdd100k.com/format.html#format-conversion

$ python3 -m bdd100k.label.to_coco -m det|box_track|pose \
    -i ${in_path} -o ${out_path} [--nproc ${process_num}]

Convert2YOLO

coco format to yolo format

## training
# python3 example.py --datasets COCO --img_path /home/jay/DataSets/bdd100k/images/100k/train --label /home/jay/DataSets/bdd100k/train.json --convert_output_path /home/jay/DataSets/bdd100k/images/100k/train --img_type ".jpg" --manifest_path ./ --cls_list_file /home/jay/DataSets/bdd100k/images/100k/bdd.names

## validation
# python3 example.py --datasets COCO --img_path /home/jay/DataSets/bdd100k/images/100k/val --label /home/jay/DataSets/bdd100k/val.json --convert_output_path /home/jay/DataSets/bdd100k/images/100k/val --img_type ".jpg" --manifest_path ./ --cls_list_file /home/jay/DataSets/bdd100k/images/100k/bdd.names