Unrealistic Feature Suppression for Generative Adversarial Networks

TL;DR

This paper introduces the Unrealistic Feature Suppression (UFS) module to enhance GAN training stability and image quality by suppressing unrealistic features, demonstrated across multiple models with improved metrics.

Contribution

The paper proposes the UFS module that maintains training stability and improves image quality in GANs, addressing limitations of sample discard methods.

Findings

UFS improves Frechet Inception Distance and Inception Score.

UFS maintains training stability across models.

Visualization shows effective suppression of unrealistic features.

Abstract

Due to the unstable nature of minimax game between generator and discriminator, improving the performance of GANs is a challenging task. Recent studies have shown that selected high-quality samples in training improve the performance of GANs. However, sampling approaches which discard samples show limitations in some aspects such as the speed of training and optimality of the networks. In this paper we propose unrealistic feature suppression (UFS) module that keeps high-quality features and suppresses unrealistic features. UFS module keeps the training stability of networks and improves the quality of generated images. We demonstrate the effectiveness of UFS module on various models such as WGAN-GP, SNGAN, and BigGAN. By using UFS module, we achieved better Frechet inception distance and inception score compared to various baseline models. We also visualize how effectively our UFS module suppresses unrealistic features through class activation maps.