jholzem/artifact_regularization_for_gans

A novel training framework for image-generating GANs which penalizes synthesis artifacts by computing a Fourier dissimilarity between synthesized and real images. Thus, it improves the detection evasiveness of GANs.

Python