Spectral Distribution Aware Image Generation

Recent advances in deep generative models for photo-realistic images have led to high quality visual results. Such learn generate data from a given training distribution such that generated can not be easily distinguished real by the human eye. Yet, recent work on detection of fake pointed out they are actually distinguishable artifacts their frequency spectra. In this paper, we propose according employing spectral discriminator. The proposed discriminator is lightweight, modular and works stably with different commonly used GAN losses. We show resulting better realistic spectra, which thus harder detect cue.