Closed leviome closed 1 year ago
Yes, it is a bug. Please revise the code by yourself.
I have updated the script: https://github.com/zju3dv/neuralbody/blob/master/zju_smpl/easymocap_to_neuralbody.py
I have updated the script: https://github.com/zju3dv/neuralbody/blob/master/zju_smpl/easymocap_to_neuralbody.py
In line 63-70, I use ZJU-Mocap and get the poses of the shape [1,69], but in neuralbody I need [1,72], I wonder do I need to transform the poses? `params = {'poses': poses, 'Rh': Rh, 'Th': Th, 'shapes': shapes}
np.save(osp.join(param_out, "{}.npy".format(root)), params)`
Hi @pengsida, I tried
easymocap_to_neuralbody.py
to make custom dataset, but I only gotvertices
. Is it a bug or just a personal misoperation?