ShamLich
ShamLich
i already set os.environ['PYOPENGL_PLATFORM'] = 'egl' but seems like only use cpu
pred_output = smpl_model(betas=pred_betas, body_pose=pred_rotmat[:, 1:], global_orient=pred_rotmat[:, [0]], pose2rot=False, transl=pred_cam_full ) points_3d = pred_output.joints() 是否需要transl=pred_cam_full这一行?保留这一行代码似乎无法正确计算loss,去掉之后就正常收敛了 我没有对spin代码给出的gt 3dkeypoint进行额外处理,若要保留transl=pred_cam_full这一行代码,需要对gt 3dkeypoint进行怎样的处理?
I attempted to follow the convert datasets code for CLIFF, but there are some issues when I try to train using those datasets, I get exponentially large MPJPE, and PA-MPJPE...