Weak6D: Weakly Supervised 6D Pose Estimation With Iterative Annotation Resolver

6D object pose estimation is an essential task in vision-based robotic grasping and manipulation. Prior works always train models with a large number of pose annotated images, limiting the efficiency of model transfer between different scenarios. This letter presents an end-to-end model named Weak6D...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	IEEE robotics and automation letters 2023-03, Vol.8 (3), p.1463-1470
Hauptverfasser:	Mu, Fengjun, Huang, Rui, Shi, Kecheng, Li, Xin, Qiu, Jing, Cheng, Hong
Format:	Artikel
Sprache:	eng
Schlagworte:	Annotations Computer vision Data models Feature extraction Grasping (robotics) Iterative methods object pose estimation Optimization Pose estimation Resolvers Semantics Training weakly-supervised learning
Online-Zugang:	Volltext bestellen
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

container_end_page	1470
container_issue	3
container_start_page	1463
container_title	IEEE robotics and automation letters
container_volume	8
creator	Mu, Fengjun Huang, Rui Shi, Kecheng Li, Xin Qiu, Jing Cheng, Hong
description	6D object pose estimation is an essential task in vision-based robotic grasping and manipulation. Prior works always train models with a large number of pose annotated images, limiting the efficiency of model transfer between different scenarios. This letter presents an end-to-end model named Weak6D , which could be learned with unannotated RGB-D data. The core of the proposed approach is the novel optimizing method Iterative Annotation Resolver, which has the ability to directly utilize the captured RGB-D data through the training process. Furthermore, we employ a weak refinement loss to optimize the pose estimation network with refined object poses. We evaluated the proposed Weak6D in the YCB-Video dataset, and experimental results show our model achieved practical results without annotated data.
doi_str_mv	10.1109/LRA.2022.3190094
format	Article
fullrecord	<record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_ieee_primary_9827570</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>9827570</ieee_id><sourcerecordid>2773450133</sourcerecordid><originalsourceid>FETCH-LOGICAL-c291t-98a6cb40f01abdff0ce9995025be5f6a96f4e68ec70133e47d2f1fa3ddf6240c3</originalsourceid><addsrcrecordid>eNpNkEFLAzEQhYMoWGrvgpeA562TZDdpvJW2aqFQqUqPId2d4Na6qcm20H_vLlvE05th3pthPkJuGQwZA_2wWI2HHDgfCqYBdHpBelwolQgl5eW_-poMYtwCAMu4EjrrkeUa7ZecPtJWdyf6dthjOJYRCyqn9NVHpLNYl9-2Ln1F12X9Sec1hqY9Ih1Xla-7yQqj3x0x3JArZ3cRB2ftk4-n2fvkJVksn-eT8SLJuWZ1okdW5psUHDC7KZyDHLXWGfBsg5mTVkuXohxhroAJgakquGPOiqJwkqeQiz657_bug_85YKzN1h9C1Zw0XCmRZm2ucUHnyoOPMaAz-9D8Ek6GgWnJmYacacmZM7kmctdFSkT8s-sRV5kC8Qta32ko</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2773450133</pqid></control><display><type>article</type><title>Weak6D: Weakly Supervised 6D Pose Estimation With Iterative Annotation Resolver</title><source>IEEE Electronic Library (IEL)</source><creator>Mu, Fengjun ; Huang, Rui ; Shi, Kecheng ; Li, Xin ; Qiu, Jing ; Cheng, Hong</creator><creatorcontrib>Mu, Fengjun ; Huang, Rui ; Shi, Kecheng ; Li, Xin ; Qiu, Jing ; Cheng, Hong</creatorcontrib><description>6D object pose estimation is an essential task in vision-based robotic grasping and manipulation. Prior works always train models with a large number of pose annotated images, limiting the efficiency of model transfer between different scenarios. This letter presents an end-to-end model named Weak6D , which could be learned with unannotated RGB-D data. The core of the proposed approach is the novel optimizing method Iterative Annotation Resolver, which has the ability to directly utilize the captured RGB-D data through the training process. Furthermore, we employ a weak refinement loss to optimize the pose estimation network with refined object poses. We evaluated the proposed Weak6D in the YCB-Video dataset, and experimental results show our model achieved practical results without annotated data.</description><identifier>ISSN: 2377-3766</identifier><identifier>EISSN: 2377-3766</identifier><identifier>DOI: 10.1109/LRA.2022.3190094</identifier><identifier>CODEN: IRALC6</identifier><language>eng</language><publisher>Piscataway: IEEE</publisher><subject>Annotations ; Computer vision ; Data models ; Feature extraction ; Grasping (robotics) ; Iterative methods ; object pose estimation ; Optimization ; Pose estimation ; Resolvers ; Semantics ; Training ; weakly-supervised learning</subject><ispartof>IEEE robotics and automation letters, 2023-03, Vol.8 (3), p.1463-1470</ispartof><rights>Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c291t-98a6cb40f01abdff0ce9995025be5f6a96f4e68ec70133e47d2f1fa3ddf6240c3</citedby><cites>FETCH-LOGICAL-c291t-98a6cb40f01abdff0ce9995025be5f6a96f4e68ec70133e47d2f1fa3ddf6240c3</cites><orcidid>0000-0001-7231-5042 ; 0000-0003-4393-4084 ; 0000-0001-8047-9610 ; 0000-0002-6153-049X ; 0000-0001-5532-9530 ; 0000-0003-3448-9492</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/9827570$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,780,784,796,27924,27925,54758</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/9827570$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Mu, Fengjun</creatorcontrib><creatorcontrib>Huang, Rui</creatorcontrib><creatorcontrib>Shi, Kecheng</creatorcontrib><creatorcontrib>Li, Xin</creatorcontrib><creatorcontrib>Qiu, Jing</creatorcontrib><creatorcontrib>Cheng, Hong</creatorcontrib><title>Weak6D: Weakly Supervised 6D Pose Estimation With Iterative Annotation Resolver</title><title>IEEE robotics and automation letters</title><addtitle>LRA</addtitle><description>6D object pose estimation is an essential task in vision-based robotic grasping and manipulation. Prior works always train models with a large number of pose annotated images, limiting the efficiency of model transfer between different scenarios. This letter presents an end-to-end model named Weak6D , which could be learned with unannotated RGB-D data. The core of the proposed approach is the novel optimizing method Iterative Annotation Resolver, which has the ability to directly utilize the captured RGB-D data through the training process. Furthermore, we employ a weak refinement loss to optimize the pose estimation network with refined object poses. We evaluated the proposed Weak6D in the YCB-Video dataset, and experimental results show our model achieved practical results without annotated data.</description><subject>Annotations</subject><subject>Computer vision</subject><subject>Data models</subject><subject>Feature extraction</subject><subject>Grasping (robotics)</subject><subject>Iterative methods</subject><subject>object pose estimation</subject><subject>Optimization</subject><subject>Pose estimation</subject><subject>Resolvers</subject><subject>Semantics</subject><subject>Training</subject><subject>weakly-supervised learning</subject><issn>2377-3766</issn><issn>2377-3766</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNpNkEFLAzEQhYMoWGrvgpeA562TZDdpvJW2aqFQqUqPId2d4Na6qcm20H_vLlvE05th3pthPkJuGQwZA_2wWI2HHDgfCqYBdHpBelwolQgl5eW_-poMYtwCAMu4EjrrkeUa7ZecPtJWdyf6dthjOJYRCyqn9NVHpLNYl9-2Ln1F12X9Sec1hqY9Ih1Xla-7yQqj3x0x3JArZ3cRB2ftk4-n2fvkJVksn-eT8SLJuWZ1okdW5psUHDC7KZyDHLXWGfBsg5mTVkuXohxhroAJgakquGPOiqJwkqeQiz657_bug_85YKzN1h9C1Zw0XCmRZm2ucUHnyoOPMaAz-9D8Ek6GgWnJmYacacmZM7kmctdFSkT8s-sRV5kC8Qta32ko</recordid><startdate>20230301</startdate><enddate>20230301</enddate><creator>Mu, Fengjun</creator><creator>Huang, Rui</creator><creator>Shi, Kecheng</creator><creator>Li, Xin</creator><creator>Qiu, Jing</creator><creator>Cheng, Hong</creator><general>IEEE</general><general>The Institute of Electrical and Electronics Engineers, Inc. (IEEE)</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SC</scope><scope>7SP</scope><scope>8FD</scope><scope>JQ2</scope><scope>L7M</scope><scope>L~C</scope><scope>L~D</scope><orcidid>https://orcid.org/0000-0001-7231-5042</orcidid><orcidid>https://orcid.org/0000-0003-4393-4084</orcidid><orcidid>https://orcid.org/0000-0001-8047-9610</orcidid><orcidid>https://orcid.org/0000-0002-6153-049X</orcidid><orcidid>https://orcid.org/0000-0001-5532-9530</orcidid><orcidid>https://orcid.org/0000-0003-3448-9492</orcidid></search><sort><creationdate>20230301</creationdate><title>Weak6D: Weakly Supervised 6D Pose Estimation With Iterative Annotation Resolver</title><author>Mu, Fengjun ; Huang, Rui ; Shi, Kecheng ; Li, Xin ; Qiu, Jing ; Cheng, Hong</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c291t-98a6cb40f01abdff0ce9995025be5f6a96f4e68ec70133e47d2f1fa3ddf6240c3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Annotations</topic><topic>Computer vision</topic><topic>Data models</topic><topic>Feature extraction</topic><topic>Grasping (robotics)</topic><topic>Iterative methods</topic><topic>object pose estimation</topic><topic>Optimization</topic><topic>Pose estimation</topic><topic>Resolvers</topic><topic>Semantics</topic><topic>Training</topic><topic>weakly-supervised learning</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Mu, Fengjun</creatorcontrib><creatorcontrib>Huang, Rui</creatorcontrib><creatorcontrib>Shi, Kecheng</creatorcontrib><creatorcontrib>Li, Xin</creatorcontrib><creatorcontrib>Qiu, Jing</creatorcontrib><creatorcontrib>Cheng, Hong</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998-Present</collection><collection>IEEE Electronic Library (IEL)</collection><collection>CrossRef</collection><collection>Computer and Information Systems Abstracts</collection><collection>Electronics & Communications Abstracts</collection><collection>Technology Research Database</collection><collection>ProQuest Computer Science Collection</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>Computer and Information Systems Abstracts Academic</collection><collection>Computer and Information Systems Abstracts Professional</collection><jtitle>IEEE robotics and automation letters</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Mu, Fengjun</au><au>Huang, Rui</au><au>Shi, Kecheng</au><au>Li, Xin</au><au>Qiu, Jing</au><au>Cheng, Hong</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Weak6D: Weakly Supervised 6D Pose Estimation With Iterative Annotation Resolver</atitle><jtitle>IEEE robotics and automation letters</jtitle><stitle>LRA</stitle><date>2023-03-01</date><risdate>2023</risdate><volume>8</volume><issue>3</issue><spage>1463</spage><epage>1470</epage><pages>1463-1470</pages><issn>2377-3766</issn><eissn>2377-3766</eissn><coden>IRALC6</coden><abstract>6D object pose estimation is an essential task in vision-based robotic grasping and manipulation. Prior works always train models with a large number of pose annotated images, limiting the efficiency of model transfer between different scenarios. This letter presents an end-to-end model named Weak6D , which could be learned with unannotated RGB-D data. The core of the proposed approach is the novel optimizing method Iterative Annotation Resolver, which has the ability to directly utilize the captured RGB-D data through the training process. Furthermore, we employ a weak refinement loss to optimize the pose estimation network with refined object poses. We evaluated the proposed Weak6D in the YCB-Video dataset, and experimental results show our model achieved practical results without annotated data.</abstract><cop>Piscataway</cop><pub>IEEE</pub><doi>10.1109/LRA.2022.3190094</doi><tpages>8</tpages><orcidid>https://orcid.org/0000-0001-7231-5042</orcidid><orcidid>https://orcid.org/0000-0003-4393-4084</orcidid><orcidid>https://orcid.org/0000-0001-8047-9610</orcidid><orcidid>https://orcid.org/0000-0002-6153-049X</orcidid><orcidid>https://orcid.org/0000-0001-5532-9530</orcidid><orcidid>https://orcid.org/0000-0003-3448-9492</orcidid></addata></record>
fulltext	fulltext_linktorsrc
identifier	ISSN: 2377-3766
ispartof	IEEE robotics and automation letters, 2023-03, Vol.8 (3), p.1463-1470
issn	2377-3766 2377-3766
language	eng
recordid	cdi_ieee_primary_9827570
source	IEEE Electronic Library (IEL)
subjects	Annotations Computer vision Data models Feature extraction Grasping (robotics) Iterative methods object pose estimation Optimization Pose estimation Resolvers Semantics Training weakly-supervised learning
title	Weak6D: Weakly Supervised 6D Pose Estimation With Iterative Annotation Resolver
url	https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-27T10%3A05%3A45IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Weak6D:%20Weakly%20Supervised%206D%20Pose%20Estimation%20With%20Iterative%20Annotation%20Resolver&rft.jtitle=IEEE%20robotics%20and%20automation%20letters&rft.au=Mu,%20Fengjun&rft.date=2023-03-01&rft.volume=8&rft.issue=3&rft.spage=1463&rft.epage=1470&rft.pages=1463-1470&rft.issn=2377-3766&rft.eissn=2377-3766&rft.coden=IRALC6&rft_id=info:doi/10.1109/LRA.2022.3190094&rft_dat=%3Cproquest_RIE%3E2773450133%3C/proquest_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2773450133&rft_id=info:pmid/&rft_ieee_id=9827570&rfr_iscdi=true