Yolo Darknet Dataset Converter

converter.py converts datasets from COCO format to YOLO Darknet format. It currently supports the following formats:

• COCO

Example

Let's say the the COCO dataset in raw format (with train, val and annotations subfolders) is in PATH_TO_COCO.

To convert this dataset to Yolo Darknet TXT format (the one used in the model zoo) you need to do the following.

For single pose estimation, run the script with the following command:

python converter.py --path PATH_TO_COCO --pose single --keypoints 13 --outputdir ./pose_estimation

For multiple pose estimation, run the script with the following command:

python converter.py --path PATH_TO_COCO --pose multi --keypoints 17 --outputdir ./pose_estimation

The converted datasets will be saved in the ./pose_estimation/single/13kpts/train|val and ./pose_estimation/multi/17kpts/train|val directories.