Artifact found. Can you analyse the problem and optimize the network architecture?

Junshk / CinCGAN-pytorch

Unofficial Implementation of "Unsupervised Image Super-Resolution using Cycle-in-Cycle Generative Adversarial Networks" in CVPR 2018.

142 stars 27 forks source link

Artifact found. Can you analyse the problem and optimize the network architecture? #12

Closed splinter21 closed 4 years ago

splinter21 commented 4 years ago

artifact

The model is trained by the default option. The reconstructed img is in the "result" folder after 400 epoch trained.

Hawk0826 commented 4 years ago

@splinter21 Hello, I'm trying to run this network, but I got the generated SR img with serious problem. I checked this img and found its value of all pixels were around 110. Have you ever changed any options or codes to avoid this problem？