Data-Efficient Instance Generation from Instance Discrimination

TL;DR

This paper introduces InsGen, a data-efficient GAN training method that uses instance discrimination to improve image synthesis quality with limited data, outperforming previous methods especially on small datasets.

Contribution

The paper proposes a novel instance discrimination-based approach for GANs that enhances data efficiency and diversity in generated images, especially with limited training data.

Findings

Outperforms state-of-the-art on 2K FFHQ images with 23.5% FID improvement

Uses instance discrimination to leverage infinite synthesized samples

Incorporates noise perturbation to enhance discriminative power

Abstract

Generative Adversarial Networks (GANs) have significantly advanced image synthesis, however, the synthesis quality drops significantly given a limited amount of training data. To improve the data efficiency of GAN training, prior work typically employs data augmentation to mitigate the overfitting of the discriminator yet still learn the discriminator with a bi-classification (i.e., real vs. fake) task. In this work, we propose a data-efficient Instance Generation (InsGen) method based on instance discrimination. Concretely, besides differentiating the real domain from the fake domain, the discriminator is required to distinguish every individual image, no matter it comes from the training set or from the generator. In this way, the discriminator can benefit from the infinite synthesized samples for training, alleviating the overfitting problem caused by insufficient training data. A noise perturbation strategy is further introduced to improve its discriminative power. Meanwhile, the learned instance discrimination capability from the discriminator is in turn exploited to encourage the generator for diverse generation. Extensive experiments demonstrate the effectiveness of our method on a variety of datasets and training settings. Noticeably, on the setting of 2K training images from the FFHQ dataset, we outperform the state-of-the-art approach with 23.5% FID improvement.