search

IDKiro / CBDNet-pytorch

Toward Convolutional Blind Denoising of Real Photograph
MIT License
172 stars 29 forks source link

is there a wrong in loss? #27

Open xiaoxiaowannote opened 2 years ago

xiaoxiaowannote commented 2 years ago

may be your torch.lt is wrong

huntkao commented 2 years ago

asym_loss = torch.mean(if_asym torch.abs(0.3 - torch.lt(gt_noise, est_noise).float()) torch.pow(est_noise - gt_noise, 2)) should be asym_loss = torch.mean(if_asym torch.abs(0.3 - torch.lt(est_noise, gt_noise).float()) torch.pow(est_noise - gt_noise, 2)) ??

MrBled commented 1 month ago

asym_loss = torch.mean(if_asym torch.abs(0.3 - torch.lt(gt_noise, est_noise).float()) torch.pow(est_noise - gt_noise, 2)) should be asym_loss = torch.mean(if_asym torch.abs(0.3 - torch.lt(est_noise, gt_noise).float()) torch.pow(est_noise - gt_noise, 2)) ??

This is an old issue but I have also confirmed this against the paper implementation.