vchoutas / smplify-x

Expressive Body Capture: 3D Hands, Face, and Body from a Single Image
https://smpl-x.is.tue.mpg.de/
Other
1.69k stars 330 forks source link

Query on SMPL Joints for human3.6m #196

Closed Dipankar1997161 closed 1 year ago

Dipankar1997161 commented 1 year ago

hello @vchoutas,

I wish to know, [ which particular set of joints ] are used for Human3.6m when constructing the smpl parameters. Since we need 24 smpl joints, Human3.6m has 32 joints in total.

The reason I wish to know is so that I can directly use the 3d ground-truth key points of H36m and develop the smpl parameters instead of using the 2d joints and images.

I found these joints from the metadata file of the h36m dataset: h36_smpl_joint_idx = [0, 1, 2, 3, 4, 6, 7, 8, 9, 11, 12, 13, 14, 15, 16, 17, 18, 19, 24, 25, 26, 27][22, 30][20, 28] joint_names = [hips, r_thigh, r_knee, r_foot, r_toe, l_thigh, l_knee, l_foot, l_toe, spine, spine1(thorax), neck, head, site, l_shoulder, l_elbo, l_wrist, l_hand, r_shoulder, r_elbo, r_wrist, r_hand,][l_wrist_end, r_wrist_end][l_thumb, r_thumb]

Also, using simplify-x, is it directly possible to use 3d keypoints without any images to generate the parameters. Thank you in advance.