How to assign the correspoding vertices of smpl with respect to skeleton points?

[Maqingyang]

I noted that you assigned different vertices of smpl in different datasets. And there is some mismatches of these assignments. For example, when I used the ground-truth moshed parameters of Human3.6m to generate smpl model, and use JOINT_REGRESSOR_TRAIN_EXTRA to regress 3d skeleton points, the MPJPE is quite high, about 32 mm. So I want to re-assign these vertices to better evaluate your great work. If this is done, I am willing to attach the new joint regressor to this repo. But I am a little unfamiliar with how this assignment process are done. Could you kindly help me out here which tool I need to assign these vertices in smpl?

[geopavlakos]

If you are interested in Human3.6M in particular, you can use the definition of joints in JOINT_REGRESSOR_H36M, which are specifically estimated to reflect the Human3.6M set of joints.

[Maqingyang]

Thank you! I will try this regressor. Furthermore, I may utilize a new dataset, MPI_INF_3DHP, to train the model. This dataset has labeled the keypoints of human toes and endpoints of hands. It may need some tools to generate a new regressor. It there any good advice?

[geopavlakos]

For this kind of locations (i.e., landmarks on the surface instead of joints in the interior of the body) it should be pretty easy to establish correspondence by manually selecting a vertex of the mesh. For the SMPL topology, the correspondences for some of the toes and hand endpoints have already been defined here.

[Maqingyang]

Thank you very very much! It seems that this file contains corresspondences of SMPL-H and SMPL-X. Are SMPL corresspondences the same as one of them?

[geopavlakos]

SMPL and SMPL+H share the same topology, so you can use the correspondences from SMPL+H.

[Maqingyang]

I will try it now！