search

hygenie1228 / ClothWild_RELEASE

[ECCV 2022] This repo is official PyTorch implementation of 3D Clothed Human Reconstruction in the Wild.
174 stars 13 forks source link

smplx parameters #7

Open wayying opened 1 year ago

wayying commented 1 year ago

To get the smpl parameters, I use whole-body codes,The final result is smpl-x param.json as follows,how do I change the parameter settings?

save SMPL-X parameters

root_pose = out['smplx_root_pose'].detach().cpu().numpy()[0] body_pose = out['smplx_body_pose'].detach().cpu().numpy()[0] lhand_pose = out['smplx_lhand_pose'].detach().cpu().numpy()[0] rhand_pose = out['smplx_rhand_pose'].detach().cpu().numpy()[0] jaw_pose = out['smplx_jaw_pose'].detach().cpu().numpy()[0] shape = out['smplx_shape'].detach().cpu().numpy()[0] expr = out['smplx_expr'].detach().cpu().numpy()[0] with open('smplx_param.json', 'w') as f: json.dump({'root_pose': root_pose.reshape(-1).tolist(), \ 'body_pose': body_pose.reshape(-1).tolist(), \ 'lhand_pose': lhand_pose.reshape(-1).tolist(), \ 'rhand_pose': rhand_pose.reshape(-1).tolist(), \ 'jaw_pose': jaw_pose.reshape(-1).tolist(), \ 'shape': shape.reshape(-1).tolist(), \ 'expr': expr.reshape(-1).tolist()}, f)

hygenie1228 commented 1 year ago

Unfortunately, ClothWild does not support SMPL-X human body model. Since SMPL and SMPL-X have different mesh topologies, the reconstruction outputs can be implausible when using SMPL-X parameters. To get the SMPL parameters, please run the [Pose2Pose model] (body only) Then, refer [this] to save the parameters. Thank you.