DragAPart: Learning a Part-Level Motion Prior for Articulated Objects

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Li, Ruining, Zheng, Chuanxia, Rupprecht, Christian, Vedaldi, Andrea
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914890279550976
author Li, Ruining
Zheng, Chuanxia
Rupprecht, Christian
Vedaldi, Andrea
author_facet Li, Ruining
Zheng, Chuanxia
Rupprecht, Christian
Vedaldi, Andrea
contents We introduce DragAPart, a method that, given an image and a set of drags as input, generates a new image of the same object that responds to the action of the drags. Differently from prior works that focused on repositioning objects, DragAPart predicts part-level interactions, such as opening and closing a drawer. We study this problem as a proxy for learning a generalist motion model, not restricted to a specific kinematic structure or object category. We start from a pre-trained image generator and fine-tune it on a new synthetic dataset, Drag-a-Move, which we introduce. Combined with a new encoding for the drags and dataset randomization, the model generalizes well to real images and different categories. Compared to prior motion-controlled generators, we demonstrate much better part-level motion understanding.
format Preprint
id arxiv_https___arxiv_org_abs_2403_15382
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle DragAPart: Learning a Part-Level Motion Prior for Articulated Objects
Li, Ruining
Zheng, Chuanxia
Rupprecht, Christian
Vedaldi, Andrea
Computer Vision and Pattern Recognition
We introduce DragAPart, a method that, given an image and a set of drags as input, generates a new image of the same object that responds to the action of the drags. Differently from prior works that focused on repositioning objects, DragAPart predicts part-level interactions, such as opening and closing a drawer. We study this problem as a proxy for learning a generalist motion model, not restricted to a specific kinematic structure or object category. We start from a pre-trained image generator and fine-tune it on a new synthetic dataset, Drag-a-Move, which we introduce. Combined with a new encoding for the drags and dataset randomization, the model generalizes well to real images and different categories. Compared to prior motion-controlled generators, we demonstrate much better part-level motion understanding.
title DragAPart: Learning a Part-Level Motion Prior for Articulated Objects
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.15382