Learning Disentangled Representations with Reference-Based Variational Autoencoders

TL;DR

This paper introduces a novel reference-based variational autoencoder framework that learns disentangled representations from unlabeled images using minimal supervision from a reference set with constant factors.

Contribution

It proposes a new weakly-supervised learning setting and a corresponding deep generative model to learn disentangled representations without explicit factor labels.

Findings

Successfully learns disentangled features with minimal supervision

Enables conditional image generation and attribute transfer

Validated on multiple tasks demonstrating effectiveness

Abstract

Learning disentangled representations from visual data, where different high-level generative factors are independently encoded, is of importance for many computer vision tasks. Solving this problem, however, typically requires to explicitly label all the factors of interest in training images. To alleviate the annotation cost, we introduce a learning setting which we refer to as "reference-based disentangling". Given a pool of unlabeled images, the goal is to learn a representation where a set of target factors are disentangled from others. The only supervision comes from an auxiliary "reference set" containing images where the factors of interest are constant. In order to address this problem, we propose reference-based variational autoencoders, a novel deep generative model designed to exploit the weak-supervision provided by the reference set. By addressing tasks such as feature learning, conditional image generation or attribute transfer, we validate the ability of the proposed model to learn disentangled representations from this minimal form of supervision.