Open vhank opened 5 years ago
The codes first down-sample the input image with the Bicubic kernel to simulate the downsampling processing. We treat the downsampled images as the inputs and the model will upsample it with the upscaling factor of 4.
You need to train/fine-tune a new model without upsampling.
So the input image should actually be a HR image, because the code will automatically use the bicubic kernel to generate the LR image. is that right?
Yes, you are right.
why the size of output is same as that of input when I run SFTGAN test code with pretrained models?