Closed chenyu1014 closed 4 months ago
This is done only during training. This is done to ensure that the model learns equally from both syn and ugc images. We observed that using only one of them was making the model biased towards those kind of distortions, so we concatenated syn and ugc features to alleviate this issue
Thank you!!!
Please tell me why we need to concat syn and ugc together syn_i1 = syn_i1.cuda(non_blocking=True) ugc_i1 = ugc_i1.cuda(non_blocking=True) x_i1 = torch.cat((syn_i1,ugc_i1),dim=0)