Can Best-Buddy Loss train without GAN and perceptual loss?

[Ken1256]

Will it better than L1 loss?
Thanks

[fenglinglwb]

Thanks for the nice comments. We only incorporate our BBL into the adversarial training, outperforming than L1 loss. You can try your proposal.

[Ken1256]

get_flat_mask
l_img = (0.2989 * r + 0.587 * g + 0.114 * b).unsqueeze(dim=1)

0.2989 0.587 0.114 where the RGB number come from?

[fenglinglwb]

It just converts a color image to a grey image...