Class Balancing GAN with a Classifier in the Loop

TL;DR

This paper introduces a class balancing regularizer for GANs that leverages a pre-trained classifier to improve generation quality on imbalanced datasets, especially long-tailed distributions.

Contribution

It proposes a novel theoretically motivated regularizer that uses classifier knowledge to balance class learning in GANs, addressing a key limitation of existing methods.

Findings

Improved FID score from 13.03 to 9.01 on iNaturalist-2019 dataset.

Demonstrated better class representation on long-tailed datasets.

Outperformed existing approaches in learning representations for imbalanced data.

Abstract

Generative Adversarial Networks (GANs) have swiftly evolved to imitate increasingly complex image distributions. However, majority of the developments focus on performance of GANs on balanced datasets. We find that the existing GANs and their training regimes which work well on balanced datasets fail to be effective in case of imbalanced (i.e. long-tailed) datasets. In this work we introduce a novel theoretically motivated Class Balancing regularizer for training GANs. Our regularizer makes use of the knowledge from a pre-trained classifier to ensure balanced learning of all the classes in the dataset. This is achieved via modelling the effective class frequency based on the exponential forgetting observed in neural networks and encouraging the GAN to focus on underrepresented classes. We demonstrate the utility of our regularizer in learning representations for long-tailed distributions via achieving better performance than existing approaches over multiple datasets. Specifically, when applied to an unconditional GAN, it improves the FID from $13.03$ to $9.01$ on the long-tailed iNaturalist-$2019$ dataset.