LBAM_Pytorch
LBAM_Pytorch copied to clipboard
code of loss part may have some problem?
hello, thank you for your excellent work and the open source code. And I have some questions about InpaintingLoss.py:
https://github.com/Vious/LBAM_Pytorch/blob/98c2ae70486f4ba3ab86d4345e586e7841cfe343/loss/InpaintingLoss.py#L75
why not use D_loss = D_fake - D_real + gp? if D_real = D_real.mean().sum() * -1 and D_fake = D_fake.mean().sum() * 1 I think D_loss = D_fake + D_real + gp
and line 120 GLoss = holeLoss + validAreaLoss + prcLoss + styleLoss + 0.1 * D_fake should be GLoss = holeLoss + validAreaLoss + prcLoss + styleLoss - 0.1 * D_fake
thank you for your reply.
same confused ...
Have you solved it?