Improved Consistency Regularization for GANs

2020/02/11 by Zhengli Zhao, Sameer Singh, Zhao, Zhengli +9 · 6 citations
Computer Science · Mathematics · #Advanced Neural Network Applications #Adversarial Robustness in Machine Learning #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.2002.04724

AAAI 2021

openalex publication_date 2020/02/11 · arxiv created 2020/12/14 · arxiv updated 2020/12/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Recent work has increased the performance of Generative Adversarial Networks (GANs) by enforcing a consistency cost on the discriminator. We improve on this technique in several ways. We first show that consistency regularization can introduce artifacts into the GAN samples and explain how to fix this issue. We then propose several modifications to the consistency regularization procedure designed to improve its performance. We carry out extensive experiments quantifying the benefit of our improvements. For unconditional image synthesis on CIFAR-10 and CelebA, our modifications yield the best known FID scores on various GAN architectures. For conditional image synthesis on CIFAR-10, we improve the state-of-the-art FID score from 11.48 to 9.21. Finally, on ImageNet-2012, we apply our technique to the original BigGAN model and improve the FID from 6.66 to 5.38, which is the best score at that model size.

Citations

• ImageNet Large Scale Visual Recognition Challenge
• Improved Techniques for Training GANs
• Conditional Image Synthesis With Auxiliary Classifier GANs
• Improved Regularization of Convolutional Neural Networks with Cutout
• Hierarchical Implicit Models and Likelihood-Free Variational Inference
• On Convergence and Stability of GANs
• Self-Attention Generative Adversarial Networks
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Temporal Ensembling for Semi-Supervised Learning
• Unsupervised Data Augmentation for Consistency Training
• Small-GAN: Speeding Up GAN Training Using Core-sets
• Image Augmentations for GAN Training
• Training Generative Adversarial Networks with Limited Data
• Learning Multiple Layers of Features from Tiny Images

Cited by

• A Turn Toward Better Alignment: Few-Shot Generative Adaptation with Equivariant Feature Rotation
• SONA: Learning Conditional, Unconditional, and Mismatching-Aware Discriminator
• APT: Adaptive Personalized Training for Diffusion Models with Limited Data
• Dual Contrastive Loss and Attention for GANs
• Learning Temporally Invariant and Localizable Features via Data Augmentation for Video Recognition
• Generative Adversarial Networks with Limited Data: A Survey and Benchmarking

Related

• Consistency Regularization for Generative Adversarial Networks
• Consistency Regularization with Generative Adversarial Networks for Semi-Supervised Learning
• Stabilizing Training of Generative Adversarial Networks through Regularization
• Feature Statistics Mixing Regularization for Generative Adversarial Networks
• Large Scale GAN Training for High Fidelity Natural Image Synthesis