Pix2Pix not translating image.
redplanet543 opened this issue · comments
I have a data set of paired images where the source image is a side or frontal view of a face and the target image is the exact same picture in the source image plus a big red dot located on the tip of the nose. I'm training Pix2Pix to perform this transformation but the resulting image is simply the source image without any modification at all. What's so hard about this transformation? Are the differences between the source and target images too small?
Not sure why. You can try to decrease the L1 loss.