yoknapathawa

Vision transformer trained on 300M human images with state of the art results on a bunch of human tasks (keypoints, segmentation, depth, normals).

Disclaimer: Co-author here.