PatchGAN

• Type of discriminator
• only penalizes structure at the scale of local image patches
• tries to classify if each $N\times N$ patch in an image is real or fake
• discriminator is run convolutionally across the image, averaging all responses to provide the ultimate output of $D$
• effectively models the image as a Markov random field
• assuming independence between pixels separated by more than a patch diameter
• type of texture/style loss
• rather the regular GAN maps from a 256×256 image to a single scalar output, which signifies “real” or “fake”, whereas the PatchGAN maps from 256×256 to an NxN (here 70×70) array of outputs X, where each $X_{ij}$ signifies whether the patch ij in the image is real or fake.
•