Self-supervised Enhancement of Latent Discovery in GANs

TL;DR

This paper introduces SRE, a self-supervised method that improves the disentanglement of latent directions in GANs, enabling better interpretability and attribute-based image retrieval.

Contribution

The paper proposes a novel self-supervised Scale Ranking Estimator (SRE) that enhances latent space disentanglement in GANs without requiring attribute labels.

Findings

SRE significantly improves disentanglement across datasets.

Enhanced directions enable attribute-based image retrieval.

Qualitative and quantitative evaluations confirm effectiveness.

Abstract

Several methods for discovering interpretable directions in the latent space of pre-trained GANs have been proposed. Latent semantics discovered by unsupervised methods are relatively less disentangled than supervised methods since they do not use pre-trained attribute classifiers. We propose Scale Ranking Estimator (SRE), which is trained using self-supervision. SRE enhances the disentanglement in directions obtained by existing unsupervised disentanglement techniques. These directions are updated to preserve the ordering of variation within each direction in latent space. Qualitative and quantitative evaluation of the discovered directions demonstrates that our proposed method significantly improves disentanglement in various datasets. We also show that the learned SRE can be used to perform Attribute-based image retrieval task without further training.