Towards Faster and Stabilized GAN Training for High-fidelity Few-shot\n Image Synthesis

2021/01/12 by Bingchen Liu, Liu, Bingchen, Yizhe Zhu +5 · 7 citations
Computer Science · #Generative Adversarial Networks and Image Synthesis #Advanced Image Processing Techniques #Digital Media Forensic Detection

paper · pdf · doi:10.48550/arxiv.2101.04775

Abstract

Training Generative Adversarial Networks (GAN) on high-fidelity images\nusually requires large-scale GPU-clusters and a vast number of training images.\nIn this paper, we study the few-shot image synthesis task for GAN with minimum\ncomputing cost. We propose a light-weight GAN structure that gains superior\nquality on 1024*1024 resolution. Notably, the model converges from scratch with\njust a few hours of training on a single RTX-2080 GPU, and has a consistent\nperformance, even with less than 100 training samples. Two technique designs\nconstitute our work, a skip-layer channel-wise excitation module and a\nself-supervised discriminator trained as a feature-encoder. With thirteen\ndatasets covering a wide variety of image domains (The datasets and code are\navailable at: https://github.com/odegeasslbc/FastGAN-pytorch), we show our\nmodel's superior performance compared to the state-of-the-art StyleGAN2, when\ndata and computing budget are limited.\n

Cited by

• StyleFusion360: View-Consistent Head Stylization via Adaptive Style Modulation
• Rethinking Bias in Generative Data Augmentation for Medical AI: a Frequency Recalibration Method
• SinSEMI: A One-Shot Image Generation Model and Data-Efficient Evaluation Framework for Semiconductor Inspection Equipment
• One Pass Is Not Enough: Recursive Latent Refinement for Generative Models
• FoilDiff: A Hybrid Transformer Backbone for Diffusion-based Modelling of 2D Airfoil Flow Fields
• Adversarial Defense in Cybersecurity: A Systematic Review of GANs for Threat Detection and Mitigation
• Unleashing Transformers: Parallel Token Prediction with Discrete Absorbing Diffusion for Fast High-Resolution Image Generation from Vector-Quantized Codes

Related

• Improving the Speed and Quality of GAN by Adversarial Training
• One-shot Ultra-high-Resolution Generative Adversarial Network That Synthesizes 16K Images On A Single GPU
• An Acceleration Framework for High Resolution Image Synthesis
• Few-Shot Adaptation of Generative Adversarial Networks
• Time Efficient Training of Progressive Generative Adversarial Network using Depthwise Separable Convolution and Super Resolution Generative Adversarial Network