Calculating PVE

[katefive]

Hi! Thanks for the great work.

I'm running on the ROMP model and trying to calculate PVE so I switched the argument calc_PVE_error to True in the config file (eval_3dpw_test_resnet.yml). But an error occurs because the dimensions of gt and output are not matched.

    outputs['meta_data']['verts'][val_idx]-outputs['verts'][val_idx], p=2, dim=-1).mean(-1)
RuntimeError: The size of tensor a (64) must match the size of tensor b (2) at non-singleton dimension 1

I checked line 66 of eval.py and found that the output only includes the mesh vertices of one person, not multiple persons.

print("gt: ", outputs['meta_data']['verts'][val_idx].shape)    # 2B, 64, 6890, 3
print("out: ", outputs['verts'][val_idx].shape) # 2B, 6890, 3

How can I fix it? And I'm also wondering why the number of the first dimension is twice the batch size.

Thank you.

[neverstoplearn]

val.py and found that the output only includes the mesh vertices of one person, not multiple persons.

I meet the same error,Have you fix it? thanks.