Hi, Xintao,
I just trained to use esrgan to train my dataset, but the SR images with quite artifacts, could you give some guidance on how to handle them?
The LR images are generated from another model but these LR images are paired with HR GT.
I tried to tune the parameters many times. I used a large parameter for the pixel-wise loss, the SR images are too blurred. When I changed to the small parameter and keep VGG parameter as 1, it is somehow not blurred, but the SR images are too noisy and with artifacts. I have no idea how to hand this problem, could you offer some help?
Hi, Xintao, I just trained to use esrgan to train my dataset, but the SR images with quite artifacts, could you give some guidance on how to handle them? The LR images are generated from another model but these LR images are paired with HR GT. I tried to tune the parameters many times. I used a large parameter for the pixel-wise loss, the SR images are too blurred. When I changed to the small parameter and keep VGG parameter as 1, it is somehow not blurred, but the SR images are too noisy and with artifacts. I have no idea how to hand this problem, could you offer some help?