DualContrast: Unsupervised Disentangling of Content and Transformations with Implicit Parameterization

TL;DR

DualContrast is a novel unsupervised contrastive method that effectively disentangles content and transformations in shape-focused images, surpassing existing explicit and contrastive approaches, demonstrated on cellular protein images.

Contribution

The paper introduces DualContrast, a new contrastive generative model for unsupervised disentanglement of content and transformations without explicit parameterization.

Findings

Outperforms existing self-supervised and explicit methods

Successfully disentangled protein composition and conformations in 3D images

Demonstrates effectiveness on shape-focused scientific datasets

Abstract

Unsupervised disentanglement of content and transformation is significantly important for analyzing shape-focused scientific image datasets, given their efficacy in solving downstream image-based shape-analyses tasks. The existing relevant works address the problem by explicitly parameterizing the transformation latent codes in a generative model, significantly reducing their expressiveness. Moreover, they are not applicable in cases where transformations can not be readily parametrized. An alternative to such explicit approaches is contrastive methods with data augmentation, which implicitly disentangles transformations and content. However, the existing contrastive strategies are insufficient to this end. Therefore, we developed a novel contrastive method with generative modeling, DualContrast, specifically for unsupervised disentanglement of content and transformations in shape-focused image datasets. DualContrast creates positive and negative pairs for content and transformation from data and latent spaces. Our extensive experiments showcase the efficacy of DualContrast over existing self-supervised and explicit parameterization approaches. With DualContrast, we disentangled protein composition and conformations in cellular 3D protein images, which was unattainable with existing disentanglement approaches