search

paTRICK-swk / D3DP

[ICCV2023] The PyTorch implementation for "Diffusion-Based 3D Human Pose Estimation with Multi-Hypothesis Aggregation"
MIT License
158 stars 8 forks source link

bug for rearrange #2

Closed zhLawliet closed 1 year ago

zhLawliet commented 1 year ago

when i want to run "python main_draw.py -k cpn_ft_h36m_dbb -b 2 -c checkpoint -gpu 0 --nolog --evaluate h36m_best_epoch.bin -num_proposals 5 -sampling_timesteps 5 --render --viz-subject S11 --viz-action SittingDown --viz-camera 1" there is a bug for inputs_2d_p = rearrange(inputs_2d_p, 'b f c -> f c b') , i find the inputs_2d_p is torch.Size([2, 2356, 17, 2]).

paTRICK-swk commented 1 year ago

Thanks for pointing out this problem. The bug has been fixed.

zhLawliet commented 1 year ago

thanks, i find you set self.augment = False, which means there is no flipped version for the model?

paTRICK-swk commented 1 year ago

The flipping augmentation will still be used, as it is implemented here.

zhLawliet commented 1 year ago

thanks