PartRM: Modeling Part-Level Dynamics with Large Cross-State Reconstruction Model

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Gao, Mingju, Pan, Yike, Gao, Huan-ang, Zhang, Zongzheng, Li, Wenyi, Dong, Hao, Tang, Hao, Yi, Li, Zhao, Hao
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912293451726848
author Gao, Mingju
Pan, Yike
Gao, Huan-ang
Zhang, Zongzheng
Li, Wenyi
Dong, Hao
Tang, Hao
Yi, Li
Zhao, Hao
author_facet Gao, Mingju
Pan, Yike
Gao, Huan-ang
Zhang, Zongzheng
Li, Wenyi
Dong, Hao
Tang, Hao
Yi, Li
Zhao, Hao
contents As interest grows in world models that predict future states from current observations and actions, accurately modeling part-level dynamics has become increasingly relevant for various applications. Existing approaches, such as Puppet-Master, rely on fine-tuning large-scale pre-trained video diffusion models, which are impractical for real-world use due to the limitations of 2D video representation and slow processing times. To overcome these challenges, we present PartRM, a novel 4D reconstruction framework that simultaneously models appearance, geometry, and part-level motion from multi-view images of a static object. PartRM builds upon large 3D Gaussian reconstruction models, leveraging their extensive knowledge of appearance and geometry in static objects. To address data scarcity in 4D, we introduce the PartDrag-4D dataset, providing multi-view observations of part-level dynamics across over 20,000 states. We enhance the model's understanding of interaction conditions with a multi-scale drag embedding module that captures dynamics at varying granularities. To prevent catastrophic forgetting during fine-tuning, we implement a two-stage training process that focuses sequentially on motion and appearance learning. Experimental results show that PartRM establishes a new state-of-the-art in part-level motion learning and can be applied in manipulation tasks in robotics. Our code, data, and models are publicly available to facilitate future research.
format Preprint
id arxiv_https___arxiv_org_abs_2503_19913
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PartRM: Modeling Part-Level Dynamics with Large Cross-State Reconstruction Model
Gao, Mingju
Pan, Yike
Gao, Huan-ang
Zhang, Zongzheng
Li, Wenyi
Dong, Hao
Tang, Hao
Yi, Li
Zhao, Hao
Computer Vision and Pattern Recognition
As interest grows in world models that predict future states from current observations and actions, accurately modeling part-level dynamics has become increasingly relevant for various applications. Existing approaches, such as Puppet-Master, rely on fine-tuning large-scale pre-trained video diffusion models, which are impractical for real-world use due to the limitations of 2D video representation and slow processing times. To overcome these challenges, we present PartRM, a novel 4D reconstruction framework that simultaneously models appearance, geometry, and part-level motion from multi-view images of a static object. PartRM builds upon large 3D Gaussian reconstruction models, leveraging their extensive knowledge of appearance and geometry in static objects. To address data scarcity in 4D, we introduce the PartDrag-4D dataset, providing multi-view observations of part-level dynamics across over 20,000 states. We enhance the model's understanding of interaction conditions with a multi-scale drag embedding module that captures dynamics at varying granularities. To prevent catastrophic forgetting during fine-tuning, we implement a two-stage training process that focuses sequentially on motion and appearance learning. Experimental results show that PartRM establishes a new state-of-the-art in part-level motion learning and can be applied in manipulation tasks in robotics. Our code, data, and models are publicly available to facilitate future research.
title PartRM: Modeling Part-Level Dynamics with Large Cross-State Reconstruction Model
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.19913