search

MIVRC / SeaNet-PyTorch

This repository is a PyTorch version of "Soft-edge Assisted Network for Single Image Super-Resolution". (IEEE TIP 2020)
27 stars 10 forks source link

loss adversarial.py #2

Closed sheepcyyx closed 3 years ago

sheepcyyx commented 3 years ago

Hello, I've been reading your code recently. I have a question about the adversarial.py. According to the adversarial.py, I used --loss 1* GaN for training, which I think it is to use discriminator to judge HR and SR. But I found that the effect was poor. I would like to know if you have done a similar experiment?Looking forward to your reply！

CV-JunchengLi commented 3 years ago

GAN is difficult to train, how to weigh the adversarial loss is also worth studying. I have not tried using the adversarial loss to train with my model, so I cannot help you.

sheepcyyx commented 3 years ago

Thanks a lot！