InfoGAN-CR and ModelCentrality: Self-supervised Model Training and Selection for Disentangling GANs

TL;DR

This paper introduces a self-supervised approach for training disentangled GANs using a contrastive regularizer based on latent traversal, and proposes an unsupervised model selection method called ModelCentrality, achieving state-of-the-art results without supervised labels.

Contribution

It presents a novel self-supervised training method for disentangled GANs and an unsupervised model selection scheme, improving disentanglement scores without requiring labeled data.

Findings

Contrastive regularizer outperforms previous methods in disentanglement scores.

ModelCentrality effectively identifies high-quality models without ground-truth labels.

Combining both methods yields significant improvements over state-of-the-art approaches.

Abstract

Disentangled generative models map a latent code vector to a target space, while enforcing that a subset of the learned latent codes are interpretable and associated with distinct properties of the target distribution. Recent advances have been dominated by Variational AutoEncoder (VAE)-based methods, while training disentangled generative adversarial networks (GANs) remains challenging. In this work, we show that the dominant challenges facing disentangled GANs can be mitigated through the use of self-supervision. We make two main contributions: first, we design a novel approach for training disentangled GANs with self-supervision. We propose contrastive regularizer, which is inspired by a natural notion of disentanglement: latent traversal. This achieves higher disentanglement scores than state-of-the-art VAE- and GAN-based approaches. Second, we propose an unsupervised model selection scheme called ModelCentrality, which uses generated synthetic samples to compute the medoid (multi-dimensional generalization of median) of a collection of models. The current common practice of hyper-parameter tuning requires using ground-truths samples, each labelled with known perfect disentangled latent codes. As real datasets are not equipped with such labels, we propose an unsupervised model selection scheme and show that it finds a model close to the best one, for both VAEs and GANs. Combining contrastive regularization with ModelCentrality, we improve upon the state-of-the-art disentanglement scores significantly, without accessing the supervised data.