High-Fidelity Image Generation With Fewer Labels

Abstract

Deep generative models are becoming a cornerstone of modern machine learning. Recent work on conditional generative adversarial networks has shown that learning complex, high-dimensional distributions over natural images is within reach. While the latest models are able to generate high-fidelity, diverse natural images at high resolution, they rely on a vast quantity of labeled data. In this work we demonstrate how one can benefit from recent work on self- and semi-supervised learning to outperform the state of the art on both unsupervised ImageNet synthesis, as well as in the conditional setting. In particular, the proposed approach is able to match the sample quality (as measured by FID) of the current state-of-the-art conditional model BigGAN on ImageNet using only 10% of the labels and outperform it using 20% of the labels.

Mario Lucic, Michael Tschannen, and Marvin Ritter contributed equally to this work. ICML 2019 camera-ready version. Code available at https://github.com/google/compare_gan

References in corpus (3)

• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• A Learned Representation For Artistic Style
• Semi-supervised Conditional GANs

Cited by in corpus (10)

• Diffusion Models Beat GANs on Image Synthesis
• Image Augmentations for GAN Training
• Text to Image Synthesis using Stacked Conditional Variational Autoencoders and Conditional Generative Adversarial Networks
• Revisiting Image Aesthetic Assessment via Self-Supervised Feature Learning
• Attentive Normalization for Conditional Image Generation
• Semi-supervised Learning using Adversarial Training with Good and Bad Samples
• Omni-GAN: On the Secrets of cGANs and Beyond
• Robust conditional GANs under missing or uncertain labels
• Class Balancing GAN with a Classifier in the Loop
• Instance-Conditioned GAN