Hi, nice work! I wonder how you evaluate the method on DIV2K dataset. Do you crop paired LR-HR patches from the original image (for example, 48x48 LR img and 192x192 HR img cropped from original 2K img), or do you resize the whole image to the target resolution (for example, resize the whole image to 48x48 and 192x192)
Hi, nice work! I wonder how you evaluate the method on DIV2K dataset. Do you crop paired LR-HR patches from the original image (for example, 48x48 LR img and 192x192 HR img cropped from original 2K img), or do you resize the whole image to the target resolution (for example, resize the whole image to 48x48 and 192x192)