DisUnknown: Distilling Unknown Factors for Disentanglement Learning

TL;DR

DisUnknown is a flexible framework that disentangles labeled and unknown factors in data, enabling controllable generation even when some factors are unlabelled, by distilling unknown factors through a two-stage training process.

Contribution

It introduces a novel weakly-supervised multi-factor disentanglement method that handles unknown factors, improving controllability in generative models without requiring full labels.

Findings

Effective disentanglement of unknown factors demonstrated on benchmarks.

Improved controllable generation with partial supervision.

Scalable to real-world complex datasets.

Abstract

Disentangling data into interpretable and independent factors is critical for controllable generation tasks. With the availability of labeled data, supervision can help enforce the separation of specific factors as expected. However, it is often expensive or even impossible to label every single factor to achieve fully-supervised disentanglement. In this paper, we adopt a general setting where all factors that are hard to label or identify are encapsulated as a single unknown factor. Under this setting, we propose a flexible weakly-supervised multi-factor disentanglement framework DisUnknown, which Distills Unknown factors for enabling multi-conditional generation regarding both labeled and unknown factors. Specifically, a two-stage training approach is adopted to first disentangle the unknown factor with an effective and robust training method, and then train the final generator with the proper disentanglement of all labeled factors utilizing the unknown distillation. To demonstrate the generalization capacity and scalability of our method, we evaluate it on multiple benchmark datasets qualitatively and quantitatively and further apply it to various real-world applications on complicated datasets.