The difference between 'joint_img' and 'mano_joint_img'

[ZYX-MLer]

InterWild/data/ReInterHand/ReInterHand.py

Line 353 in 0c4cd6c

targets = {'joint_img': joint_img, 'mano_joint_img': mano_joint_img, 'joint_cam': joint_cam, 'mano_mesh_cam': mano_mesh_cam, 'rel_trans': rel_trans, 'mano_pose': mano_pose, 'mano_shape': mano_shape, 'lhand_bbox_center': lhand_bbox_center, 'lhand_bbox_size': lhand_bbox_size, 'rhand_bbox_center': rhand_bbox_center, 'rhand_bbox_size': rhand_bbox_size}

What is the difference between 'joint_img' and 'mano_joint_img'?
Which one should I use for hand keypoint detection？

thank you.

[mks0601]

joint_img is from dataset groundtruth 2.5D keypoints.
mano_joint_img is from projecting groundtruth MANO 3D keypoints to the 2.5D space.
As in the script, I suggest using both.