# Disentangling Content and Style via Unsupervised Geometry Distillation

arXiv: https://arxiv.org/abs/1905.04538 · HTML: https://mlsanity.com/paper/1905.04538/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 11 May 2019.

## Built on by (most cited first, 9 of 9)

- [StyleDiffusion: Controllable Disentangled Style Transfer via Diffusion Models](https://mlsanity.com/paper/2308.07863/) (2023)
- [Simple Disentanglement of Style and Content in Visual Representations](https://mlsanity.com/paper/2302.09795/) (2023)
- [Generating Novel Scene Compositions from Single Images and Videos](https://mlsanity.com/paper/2103.13389/) (2021)
- [Rethinking Content and Style: Exploring Bias for Unsupervised Disentanglement](https://mlsanity.com/paper/2102.10544/) (2021)
- [Retriever: Learning Content-Style Representation as a Token-Level Bipartite Graph](https://mlsanity.com/paper/2202.12307/) (2022)
- [Harnessing the Conditioning Sensorium for Improved Image Translation](https://mlsanity.com/paper/2110.06443/) (2021)
- [Learning Geometrically Disentangled Representations of Protein Folding Simulations](https://mlsanity.com/paper/2205.10423/) (2022)
- [DualVAE: Controlling Colours of Generated and Real Images](https://mlsanity.com/paper/2305.18769/) (2023)
- [Task-Relevant Representation Decoupling for Visual Reinforcement Learning Generalization](https://mlsanity.com/paper/2607.00796/) (2026)

## Abstract (arXiv)

It is challenging to disentangle an object into two orthogonal spaces of content and style since each can influence the visual observation differently and unpredictably. It is rare for one to have access to a large number of data to help separate the influences. In this paper, we present a novel framework to learn this disentangled representation in a completely unsupervised manner. We address this problem in a two-branch Autoencoder framework. For the structural content branch, we project the latent factor into a soft structured point tensor and constrain it with losses derived from prior knowledge. This constraint encourages the branch to distill geometry information. Another branch learns the complementary style information. The two branches form an effective framework that can disentangle object's content-style representation without any human annotation. We evaluate our approach on four image datasets, on which we demonstrate the superior disentanglement and visual analogy quality both in synthesized and real-world data. We are able to generate photo-realistic images with 256*256 resolution that are clearly disentangled in content and style.