socioopticon material / spec - any other details....

[johndpope]

at the moment there's only this https://raw.githubusercontent.com/facebookresearch/sapiens/main/pose/configs/_base_/datasets/goliath.py

is there any other material on this format? whitepaper? i can't find anything - but interested in updates on this....

eg. can the model downgrade to just coco-wholebody pose?

 coco_wholebody = {
        'nose': sapien_pose['center_of_nose_root'],
        'left_shoulder': sapien_pose['left_shoulder'],
        'right_shoulder': sapien_pose['right_shoulder'],
        # ... map other body keypoints ...
        'left_wrist': sapien_pose['left_wrist'],
        'right_wrist': sapien_pose['right_wrist'],
        # ... map hand keypoints ...
    }
    return coco_wholebody

[rawalkhirodkar]

@johndpope Apart from the code reference you shared and the published paper, we do not have another details on the goliath format.

If you enlist what other information you would like, we can work on providing it.
Keypoint definitions and class segmentation labels are shared.