Issue about the difference between result when training (64x64) vs result when testing (full image) #33
Description
i fine-tuned WeatherDiff64.pth.tar on O-Haze dataset on image dehazing task.
while training, there is a validation step, which generates 64x64 patch image, and the image is very good. (image above)
but when i run eval, the image is the same as hazy image. i dont know why?