HUMOS: Human Motion Model Conditioned on Body Shape

TL;DR

This paper introduces HUMOS, a generative model that creates diverse, realistic human motions conditioned on individual body shapes, addressing limitations of previous models that used average body types.

Contribution

HUMOS is the first model to generate human motion conditioned on body shape using unpaired data and cycle consistency, improving realism and diversity.

Findings

Produces more realistic and diverse motions than previous models

Maintains physical plausibility and stability in generated motions

Effectively learns from unpaired data using cycle consistency

Abstract

Generating realistic human motion is essential for many computer vision and graphics applications. The wide variety of human body shapes and sizes greatly impacts how people move. However, most existing motion models ignore these differences, relying on a standardized, average body. This leads to uniform motion across different body types, where movements don't match their physical characteristics, limiting diversity. To solve this, we introduce a new approach to develop a generative motion model based on body shape. We show that it's possible to train this model using unpaired data by applying cycle consistency, intuitive physics, and stability constraints, which capture the relationship between identity and movement. The resulting model generates diverse, physically plausible, and dynamically stable human motions that are both quantitatively and qualitatively more realistic than current state-of-the-art methods. More details are available on our project page https://CarstenEpic.github.io/humos/.