Does output image size need to be exactly same as input image size?
Posted by: rajatguptakgp @ June 27, 2020, 7:11 a.m.I am a little bit of confused. The training data of LR/HR are of size 194x194/776x776, of which the HR patch is x4 larger than LR. However, in the validation phase, the output size is required to be the same as LR image. For example, if the LR is 1024x2048, then the HR is 1024x2048 as well. But my model is designed to only be able to handle the situation where HR is x4 of LR. May I ask how I can get the HR image the same of LR, as it is required?
Posted by: lhuo @ July 3, 2020, 1:40 p.m.Sorry, my mistake in the previous answer.
The output needs to be the 4x upscaled input LR image.
Please read the description on the challenge and evaluation protocol.