search

facebookresearch / InterHand2.6M

Official PyTorch implementation of "InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image", ECCV 2020
Other
676 stars 92 forks source link

about 'J_regressor_mano_ih26m.npy' #92

Closed nitba closed 2 years ago

nitba commented 2 years ago

Hi @mks0601

Thanks for your great work, Is J_regressor_mano_ih26m.npy identical to a joint regressor in smplx.MANO ? https://github.com/facebookresearch/InterHand2.6M/blob/2b8061d2c8e762aa6fcb8e6f5d18f8a9e83bfd0c/tool/MANO_world_to_camera/convert.py#L33

mks0601 commented 2 years ago

The joint regressor in MANO does not produce fingertips, so I manually added them. Also, the joint order is different.

nitba commented 2 years ago

I have used manopth https://github.com/hassony2/manopth which has assumed regressing the fingertips too and then I reordered the joints. I was wondering would that be identical to your regressor (I mean weights to interpolate joints)?

mks0601 commented 2 years ago

Weights are the same, but the joint order is different.