Fourier Spectrum Discrepancies in Deep Network Generated Images

TL;DR

This paper identifies a systematic discrepancy in high-frequency Fourier modes between real and deep network generated images, proposing a spectrum-based detection method with high accuracy and analyzing the effects of image transformations.

Contribution

It introduces a novel Fourier spectrum analysis for detecting deep network generated images and demonstrates high accuracy with minimal training data.

Findings

Fourier spectrum discrepancies are consistent in generated images.

The detection method achieves up to 99.2% accuracy.

Image transformations affect detection accuracy.

Abstract

Advancements in deep generative models such as generative adversarial networks and variational autoencoders have resulted in the ability to generate realistic images that are visually indistinguishable from real images, which raises concerns about their potential malicious usage. In this paper, we present an analysis of the high-frequency Fourier modes of real and deep network generated images and show that deep network generated images share an observable, systematic shortcoming in replicating the attributes of these high-frequency modes. Using this, we propose a detection method based on the frequency spectrum of the images which is able to achieve an accuracy of up to 99.2% in classifying real and deep network generated images from various GAN and VAE architectures on a dataset of 5000 images with as few as 8 training examples. Furthermore, we show the impact of image transformations such as compression, cropping, and resolution reduction on the classification accuracy and suggest a method for modifying the high-frequency attributes of deep network generated images to mimic real images.