DexGraspNet 2.0: Learning Generative Dexterous Grasping in Large-scale Synthetic Cluttered Scenes

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Jialiang, Liu, Haoran, Li, Danshi, Yu, Xinqiang, Geng, Haoran, Ding, Yufei, Chen, Jiayi, Wang, He
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912096045760512
author Zhang, Jialiang
Liu, Haoran
Li, Danshi
Yu, Xinqiang
Geng, Haoran
Ding, Yufei
Chen, Jiayi
Wang, He
author_facet Zhang, Jialiang
Liu, Haoran
Li, Danshi
Yu, Xinqiang
Geng, Haoran
Ding, Yufei
Chen, Jiayi
Wang, He
contents Grasping in cluttered scenes remains highly challenging for dexterous hands due to the scarcity of data. To address this problem, we present a large-scale synthetic benchmark, encompassing 1319 objects, 8270 scenes, and 427 million grasps. Beyond benchmarking, we also propose a novel two-stage grasping method that learns efficiently from data by using a diffusion model that conditions on local geometry. Our proposed generative method outperforms all baselines in simulation experiments. Furthermore, with the aid of test-time-depth restoration, our method demonstrates zero-shot sim-to-real transfer, attaining 90.7% real-world dexterous grasping success rate in cluttered scenes.
format Preprint
id arxiv_https___arxiv_org_abs_2410_23004
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle DexGraspNet 2.0: Learning Generative Dexterous Grasping in Large-scale Synthetic Cluttered Scenes
Zhang, Jialiang
Liu, Haoran
Li, Danshi
Yu, Xinqiang
Geng, Haoran
Ding, Yufei
Chen, Jiayi
Wang, He
Robotics
Computer Vision and Pattern Recognition
Grasping in cluttered scenes remains highly challenging for dexterous hands due to the scarcity of data. To address this problem, we present a large-scale synthetic benchmark, encompassing 1319 objects, 8270 scenes, and 427 million grasps. Beyond benchmarking, we also propose a novel two-stage grasping method that learns efficiently from data by using a diffusion model that conditions on local geometry. Our proposed generative method outperforms all baselines in simulation experiments. Furthermore, with the aid of test-time-depth restoration, our method demonstrates zero-shot sim-to-real transfer, attaining 90.7% real-world dexterous grasping success rate in cluttered scenes.
title DexGraspNet 2.0: Learning Generative Dexterous Grasping in Large-scale Synthetic Cluttered Scenes
topic Robotics
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.23004