Question about the output of the network
Naclzno opened this issue · 0 comments
Naclzno commented
Thank you for your contribution. The forward function outputs z+x, representing the clean image, where z acts as the compensatory element for the noise, akin to its negative value. For semantic segmentation tasks, should I solely utilize the feature map z without needing to combine it with x to form z+x? Is my understanding correct?