Why not diffusion the SMPLH parameters to generate SMPLH meshes #259
Description
Most papers that use diffusion models to obtain human pose seem to get the main joint points or directly diffuse the point cloud coordinates of human pose. Why doesn't anyone directly diffuse the parameters used to generate the human mesh? Taking the SMPLH model as an example, we need to obtain the main parameters such as betas, body_pose, left_hand_pose, and right_hand_pose. Or is it just that I haven't seen them; are there related papers?