How to make the style stronger?

[xuanandsix]

The following are input image, my training output from pair label supervision, and the output from your test model。 I trained my model (Super-Resolution model) on the images from your model outputs, I find it difficult to change the facial features。 Like the eyes and face texture are changed, how to do it ? I use L1Loss (weight is 1) + PerceptualLoss (weight is 1)+ GANLoss (weight is 0.1),

[Sxela]

Hi! What dataset are you training on? And what perceptual loss you are using? (and what model) Usually l1 should be a lot weaker (like 0.1 or less) than perceptual loss. You should be getting decent features even without a GAN