RGB-based Category-level Object Pose Estimation via Decoupled Metric Scale Recovery

doi:10.1109/ICRA57147.2024.10611723

	RGB-based Category-level Object Pose Estimation via Decoupled Metric Scale Recovery
	Wei, Jiaxin1,2 ; Song, Xibin 3; Liu, Weizhe 3; Kneip, Laurent2 ; Li, Hongdong 4; Ji, Pan 3
	2024
会议录名称	PROCEEDINGS - IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION
ISSN	1050-4729
页码	2036-2042
发表状态	已发表
DOI	10.1109/ICRA57147.2024.10611723
摘要	While showing promising results, recent RGB-D camera-based category-level object pose estimation methods have restricted applications due to the heavy reliance on depth sensors. RGB-only methods provide an alternative to this problem yet suffer from inherent scale ambiguity stemming from monocular observations. In this paper, we propose a novel pipeline that decouples the 6D pose and size estimation to mitigate the influence of imperfect scales on rigid transformations. Specifically, we leverage a pre-trained monocular estimator to extract local geometric information, mainly facilitating the search for inlier 2D-3D correspondence. Meanwhile, a separate branch is designed to directly recover the metric scale of the object based on category-level statistics. Finally, we advocate using the RANSAC-PnP algorithm to robustly solve for 6D object pose. Extensive experiments have been conducted on both synthetic and real datasets, demonstrating the superior performance of our method over previous state-of-the-art RGB-based approaches, especially in terms of rotation accuracy. Code: https://github.com/goldoak/DMSR. © 2024 IEEE.
会议录编者/会议主办者	Beijing NOKOV Science and Technology Co., Ltd. ; et al. ; Kawasaki Heavy Industries, Ltd. ; Kuka AG ; Schunk SE and Co. KG ; ShangHai CHINGMU Tech Ltd
关键词	Object detection RGB color model Size determination Camera-based Depth sensors Estimation methods Geometric information Object based Object pose Objects-based Pose-estimation Rigid transformations Size estimation
会议名称	2024 IEEE International Conference on Robotics and Automation, ICRA 2024
会议地点	1-1-1, Minato Mirai, Nishi-ku, Yokohama, Japan
会议日期	May 13, 2024 - May 17, 2024
URL	查看原文
收录类别	EI
语种	英语
出版者	Institute of Electrical and Electronics Engineers Inc.
EI入藏号	20243516964061
EI主题词	Object recognition
EI分类号	1106.3.1 ; 1106.8 ; 214 ; 741.1/Optics ; 941.5¬≠
原始文献类型	Conference article (CA)
来源库	IEEE
引用统计	正在获取...
文献类型	会议论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/421454
专题	信息科学与技术学院_硕士生信息科学与技术学院_PI研究组_Laurent Kneip组
通讯作者	Wei, Jiaxin
作者单位	1.Technical University of Munich, Smart Robotics Lab, CIT, Germany; 2.ShanghaiTech University, Mobile Perception Lab, SIST, China; 3.Tencent, China; 4.Australian National University, Australia
第一作者单位	上海科技大学
通讯作者单位	上海科技大学
推荐引用方式 GB/T 7714	Wei, Jiaxin,Song, Xibin,Liu, Weizhe,et al. RGB-based Category-level Object Pose Estimation via Decoupled Metric Scale Recovery[C]//Beijing NOKOV Science and Technology Co., Ltd., et al., Kawasaki Heavy Industries, Ltd., Kuka AG, Schunk SE and Co. KG, ShangHai CHINGMU Tech Ltd:Institute of Electrical and Electronics Engineers Inc.,2024:2036-2042.