Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Ziyin, Xu, Sirui, Guo, Chuan, Zhou, Bing, Gong, Jiangshan, Wang, Jian, Wang, Yu-Xiong, Gui, Liang-Yan
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.25734
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908915884621824
author Wang, Ziyin
Xu, Sirui
Guo, Chuan
Zhou, Bing
Gong, Jiangshan
Wang, Jian
Wang, Yu-Xiong
Gui, Liang-Yan
author_facet Wang, Ziyin
Xu, Sirui
Guo, Chuan
Zhou, Bing
Gong, Jiangshan
Wang, Jian
Wang, Yu-Xiong
Gui, Liang-Yan
contents Generating realistic human-object interaction (HOI) animations remains challenging because it requires jointly modeling dynamic human actions and diverse object geometries. Prior diffusion-based approaches often rely on hand-crafted contact priors or human-imposed kinematic constraints to improve contact quality. We propose LIGHT, a data-driven alternative in which guidance emerges from the denoising pace itself, reducing dependence on manually designed priors. Building on diffusion forcing, we factor the representation into modality-specific components and assign individualized noise levels with asynchronous denoising schedules. In this paradigm, cleaner components guide noisier ones through cross-attention, yielding guidance without auxiliary classifiers. We find that this data-driven guidance is inherently contact-aware, and can be enhanced when training is augmented with a broad spectrum of synthetic object geometries, encouraging invariance of contact semantics to geometric diversity. Extensive experiments show that pace-induced guidance more effectively mirrors the benefits of contact priors than conventional classifier-free guidance, while achieving higher contact fidelity, more realistic HOI generation, and stronger generalization to unseen objects and tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2603_25734
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Unleashing Guidance Without Classifiers for Human-Object Interaction Animation
Wang, Ziyin
Xu, Sirui
Guo, Chuan
Zhou, Bing
Gong, Jiangshan
Wang, Jian
Wang, Yu-Xiong
Gui, Liang-Yan
Computer Vision and Pattern Recognition
Generating realistic human-object interaction (HOI) animations remains challenging because it requires jointly modeling dynamic human actions and diverse object geometries. Prior diffusion-based approaches often rely on hand-crafted contact priors or human-imposed kinematic constraints to improve contact quality. We propose LIGHT, a data-driven alternative in which guidance emerges from the denoising pace itself, reducing dependence on manually designed priors. Building on diffusion forcing, we factor the representation into modality-specific components and assign individualized noise levels with asynchronous denoising schedules. In this paradigm, cleaner components guide noisier ones through cross-attention, yielding guidance without auxiliary classifiers. We find that this data-driven guidance is inherently contact-aware, and can be enhanced when training is augmented with a broad spectrum of synthetic object geometries, encouraging invariance of contact semantics to geometric diversity. Extensive experiments show that pace-induced guidance more effectively mirrors the benefits of contact priors than conventional classifier-free guidance, while achieving higher contact fidelity, more realistic HOI generation, and stronger generalization to unseen objects and tasks.
title Unleashing Guidance Without Classifiers for Human-Object Interaction Animation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.25734