Deforming Autoencoders: Unsupervised Disentangling of Shape and Appearance

TL;DR

Deforming Autoencoders are a novel generative model that unsupervisedly disentangles shape and appearance in images by representing shape as deformations from a canonical template, enabling advanced image manipulation and alignment.

Contribution

This work introduces a new autoencoder-based approach for unsupervised disentangling of shape and appearance using deformable templates, extending the capabilities of generative models.

Findings

Effective unsupervised group-wise image alignment.

Successful shape and appearance interpolation in various domains.

Decomposition of face images into shading and albedo.

Abstract

In this work we introduce Deforming Autoencoders, a generative model for images that disentangles shape from appearance in an unsupervised manner. As in the deformable template paradigm, shape is represented as a deformation between a canonical coordinate system (`template') and an observed image, while appearance is modeled in `canonical', template, coordinates, thus discarding variability due to deformations. We introduce novel techniques that allow this approach to be deployed in the setting of autoencoders and show that this method can be used for unsupervised group-wise image alignment. We show experiments with expression morphing in humans, hands, and digits, face manipulation, such as shape and appearance interpolation, as well as unsupervised landmark localization. A more powerful form of unsupervised disentangling becomes possible in template coordinates, allowing us to successfully decompose face images into shading and albedo, and further manipulate face images.