Saved in:
Bibliographic Details
Main Authors: Zhang, Joy Xiaoji, Zhu, Jingsen, Chen, Hanyu, Marschner, Steve
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.12600
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916847733964800
author Zhang, Joy Xiaoji
Zhu, Jingsen
Chen, Hanyu
Marschner, Steve
author_facet Zhang, Joy Xiaoji
Zhu, Jingsen
Chen, Hanyu
Marschner, Steve
contents Simulating hair dynamics that generalize across arbitrary hairstyles, body shapes, and motions is a critical challenge. Our novel two-stage neural solution is the first to leverage Transformer-based architectures for such a broad generalization. We propose a Transformer-powered static network that predicts static draped shapes for any hairstyle, effectively resolving hair-body penetrations and preserving hair fidelity. Subsequently, a dynamic network with a novel cross-attention mechanism fuses static hair features with kinematic input to generate expressive dynamics and complex secondary motions. This dynamic network also allows for efficient fine-tuning of challenging motion sequences, such as abrupt head movements. Our method offers real-time inference for both static single-frame drapes and dynamic drapes over pose sequences. Our method demonstrates high-fidelity and generalizable dynamic hair across various styles, guided by physics-informed losses, and can resolve penetrations even for complex, unseen long hairstyles, highlighting its broad generalization.
format Preprint
id arxiv_https___arxiv_org_abs_2507_12600
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HairFormer: Transformer-Based Dynamic Neural Hair Simulation
Zhang, Joy Xiaoji
Zhu, Jingsen
Chen, Hanyu
Marschner, Steve
Graphics
Computer Vision and Pattern Recognition
Simulating hair dynamics that generalize across arbitrary hairstyles, body shapes, and motions is a critical challenge. Our novel two-stage neural solution is the first to leverage Transformer-based architectures for such a broad generalization. We propose a Transformer-powered static network that predicts static draped shapes for any hairstyle, effectively resolving hair-body penetrations and preserving hair fidelity. Subsequently, a dynamic network with a novel cross-attention mechanism fuses static hair features with kinematic input to generate expressive dynamics and complex secondary motions. This dynamic network also allows for efficient fine-tuning of challenging motion sequences, such as abrupt head movements. Our method offers real-time inference for both static single-frame drapes and dynamic drapes over pose sequences. Our method demonstrates high-fidelity and generalizable dynamic hair across various styles, guided by physics-informed losses, and can resolve penetrations even for complex, unseen long hairstyles, highlighting its broad generalization.
title HairFormer: Transformer-Based Dynamic Neural Hair Simulation
topic Graphics
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2507.12600