Generative Human Geometry Distribution
Xiangjun Tang, Biao Zhang, Peter Wonka · 2025
Read Generative Human Geometry Distribution by Xiangjun Tang, Biao Zhang, Peter Wonka (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-01448.txt.
Description
Xiangjun Tang, Biao Zhang, and Peter Wonka generate varied human body geometry with clothing, in work from 2025. Realistic human shapes must keep fine clothing detail and true cloth body contact. Past geometry distributions model one person well but scale poorly to full datasets for learning.
Two changes fix that scaling gap in the new model design. Distributions are coded as flat feature maps instead of network weights for each person. A body model domain replaces a plain Gaussian base, with a tuned flow field on top. Training runs in two stages: squeeze shapes to a latent space, then learn that space. Pose based avatar tests beat prior methods with a large quality gain.
Catalog record
| CREATOR | Xiangjun Tang, Biao Zhang, Peter Wonka |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-01448