MMGSD: Multi-Modal Gaussian Shape Descriptors for Correspondence Matching in 1D and 2D Deformable Objects

We explore learning pixelwise correspondences between images of deformable objects in different configurations. Traditional correspondence matching approaches such as SIFT, SURF, and ORB can fail to provide sufficient contextual information for fine-grained manipulation. We propose Multi-Modal Gauss...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Ganapathi, Aditya, Sundaresan, Priya, Thananjeyan, Brijen, Balakrishna, Ashwin, Seita, Daniel, Hoque, Ryan, Gonzalez, Joseph E, Goldberg, Ken
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Ganapathi, Aditya
Sundaresan, Priya
Thananjeyan, Brijen
Balakrishna, Ashwin
Seita, Daniel
Hoque, Ryan
Gonzalez, Joseph E
Goldberg, Ken
description We explore learning pixelwise correspondences between images of deformable objects in different configurations. Traditional correspondence matching approaches such as SIFT, SURF, and ORB can fail to provide sufficient contextual information for fine-grained manipulation. We propose Multi-Modal Gaussian Shape Descriptor (MMGSD), a new visual representation of deformable objects which extends ideas from dense object descriptors to predict all symmetric correspondences between different object configurations. MMGSD is learned in a self-supervised manner from synthetic data and produces correspondence heatmaps with measurable uncertainty. In simulation, experiments suggest that MMGSD can achieve an RMSE of 32.4 and 31.3 for square cloth and braided synthetic nylon rope respectively. The results demonstrate an average of 47.7% improvement over a provided baseline based on contrastive learning, symmetric pixel-wise contrastive loss (SPCL), as opposed to MMGSD which enforces distributional continuity.
doi_str_mv 10.48550/arxiv.2010.04339
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2010_04339</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2010_04339</sourcerecordid><originalsourceid>FETCH-LOGICAL-a679-163eb294c5938fc95b601720f0ec92d920eb9095ca2e7da74415d7b45cb6f203</originalsourceid><addsrcrecordid>eNotj8tOwzAURL1hgQofwIr7AymOH0nNDiVQkBp1EfbR9SPUKHUiO0Xw94SW1UijOSMdQu5yuhYbKekDxm__tWZ0KajgXF0T3zTbtn6E5jTMPmtGiwNs8ZSSxwDtAScHtUsm-mkeY4J-jFCNMbo0jcG6YBw0OJuDDx_gA-Q1YLDA6gVapkfUg4O9_nRmTjfkqschudv_XJH25fm9es12--1b9bTLsChVlhfcaaaEkYpveqOkLmheMtpTZxSzilGnFVXSIHOlxVKIXNpSC2l00TPKV-T-8npW7abojxh_uj_l7qzMfwFu9VBf</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>MMGSD: Multi-Modal Gaussian Shape Descriptors for Correspondence Matching in 1D and 2D Deformable Objects</title><source>arXiv.org</source><creator>Ganapathi, Aditya ; Sundaresan, Priya ; Thananjeyan, Brijen ; Balakrishna, Ashwin ; Seita, Daniel ; Hoque, Ryan ; Gonzalez, Joseph E ; Goldberg, Ken</creator><creatorcontrib>Ganapathi, Aditya ; Sundaresan, Priya ; Thananjeyan, Brijen ; Balakrishna, Ashwin ; Seita, Daniel ; Hoque, Ryan ; Gonzalez, Joseph E ; Goldberg, Ken</creatorcontrib><description>We explore learning pixelwise correspondences between images of deformable objects in different configurations. Traditional correspondence matching approaches such as SIFT, SURF, and ORB can fail to provide sufficient contextual information for fine-grained manipulation. We propose Multi-Modal Gaussian Shape Descriptor (MMGSD), a new visual representation of deformable objects which extends ideas from dense object descriptors to predict all symmetric correspondences between different object configurations. MMGSD is learned in a self-supervised manner from synthetic data and produces correspondence heatmaps with measurable uncertainty. In simulation, experiments suggest that MMGSD can achieve an RMSE of 32.4 and 31.3 for square cloth and braided synthetic nylon rope respectively. The results demonstrate an average of 47.7% improvement over a provided baseline based on contrastive learning, symmetric pixel-wise contrastive loss (SPCL), as opposed to MMGSD which enforces distributional continuity.</description><identifier>DOI: 10.48550/arxiv.2010.04339</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Robotics</subject><creationdate>2020-10</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2010.04339$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2010.04339$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Ganapathi, Aditya</creatorcontrib><creatorcontrib>Sundaresan, Priya</creatorcontrib><creatorcontrib>Thananjeyan, Brijen</creatorcontrib><creatorcontrib>Balakrishna, Ashwin</creatorcontrib><creatorcontrib>Seita, Daniel</creatorcontrib><creatorcontrib>Hoque, Ryan</creatorcontrib><creatorcontrib>Gonzalez, Joseph E</creatorcontrib><creatorcontrib>Goldberg, Ken</creatorcontrib><title>MMGSD: Multi-Modal Gaussian Shape Descriptors for Correspondence Matching in 1D and 2D Deformable Objects</title><description>We explore learning pixelwise correspondences between images of deformable objects in different configurations. Traditional correspondence matching approaches such as SIFT, SURF, and ORB can fail to provide sufficient contextual information for fine-grained manipulation. We propose Multi-Modal Gaussian Shape Descriptor (MMGSD), a new visual representation of deformable objects which extends ideas from dense object descriptors to predict all symmetric correspondences between different object configurations. MMGSD is learned in a self-supervised manner from synthetic data and produces correspondence heatmaps with measurable uncertainty. In simulation, experiments suggest that MMGSD can achieve an RMSE of 32.4 and 31.3 for square cloth and braided synthetic nylon rope respectively. The results demonstrate an average of 47.7% improvement over a provided baseline based on contrastive learning, symmetric pixel-wise contrastive loss (SPCL), as opposed to MMGSD which enforces distributional continuity.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Robotics</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAURL1hgQofwIr7AymOH0nNDiVQkBp1EfbR9SPUKHUiO0Xw94SW1UijOSMdQu5yuhYbKekDxm__tWZ0KajgXF0T3zTbtn6E5jTMPmtGiwNs8ZSSxwDtAScHtUsm-mkeY4J-jFCNMbo0jcG6YBw0OJuDDx_gA-Q1YLDA6gVapkfUg4O9_nRmTjfkqschudv_XJH25fm9es12--1b9bTLsChVlhfcaaaEkYpveqOkLmheMtpTZxSzilGnFVXSIHOlxVKIXNpSC2l00TPKV-T-8npW7abojxh_uj_l7qzMfwFu9VBf</recordid><startdate>20201008</startdate><enddate>20201008</enddate><creator>Ganapathi, Aditya</creator><creator>Sundaresan, Priya</creator><creator>Thananjeyan, Brijen</creator><creator>Balakrishna, Ashwin</creator><creator>Seita, Daniel</creator><creator>Hoque, Ryan</creator><creator>Gonzalez, Joseph E</creator><creator>Goldberg, Ken</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20201008</creationdate><title>MMGSD: Multi-Modal Gaussian Shape Descriptors for Correspondence Matching in 1D and 2D Deformable Objects</title><author>Ganapathi, Aditya ; Sundaresan, Priya ; Thananjeyan, Brijen ; Balakrishna, Ashwin ; Seita, Daniel ; Hoque, Ryan ; Gonzalez, Joseph E ; Goldberg, Ken</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a679-163eb294c5938fc95b601720f0ec92d920eb9095ca2e7da74415d7b45cb6f203</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Robotics</topic><toplevel>online_resources</toplevel><creatorcontrib>Ganapathi, Aditya</creatorcontrib><creatorcontrib>Sundaresan, Priya</creatorcontrib><creatorcontrib>Thananjeyan, Brijen</creatorcontrib><creatorcontrib>Balakrishna, Ashwin</creatorcontrib><creatorcontrib>Seita, Daniel</creatorcontrib><creatorcontrib>Hoque, Ryan</creatorcontrib><creatorcontrib>Gonzalez, Joseph E</creatorcontrib><creatorcontrib>Goldberg, Ken</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Ganapathi, Aditya</au><au>Sundaresan, Priya</au><au>Thananjeyan, Brijen</au><au>Balakrishna, Ashwin</au><au>Seita, Daniel</au><au>Hoque, Ryan</au><au>Gonzalez, Joseph E</au><au>Goldberg, Ken</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>MMGSD: Multi-Modal Gaussian Shape Descriptors for Correspondence Matching in 1D and 2D Deformable Objects</atitle><date>2020-10-08</date><risdate>2020</risdate><abstract>We explore learning pixelwise correspondences between images of deformable objects in different configurations. Traditional correspondence matching approaches such as SIFT, SURF, and ORB can fail to provide sufficient contextual information for fine-grained manipulation. We propose Multi-Modal Gaussian Shape Descriptor (MMGSD), a new visual representation of deformable objects which extends ideas from dense object descriptors to predict all symmetric correspondences between different object configurations. MMGSD is learned in a self-supervised manner from synthetic data and produces correspondence heatmaps with measurable uncertainty. In simulation, experiments suggest that MMGSD can achieve an RMSE of 32.4 and 31.3 for square cloth and braided synthetic nylon rope respectively. The results demonstrate an average of 47.7% improvement over a provided baseline based on contrastive learning, symmetric pixel-wise contrastive loss (SPCL), as opposed to MMGSD which enforces distributional continuity.</abstract><doi>10.48550/arxiv.2010.04339</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2010.04339
ispartof
issn
language eng
recordid cdi_arxiv_primary_2010_04339
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
Computer Science - Robotics
title MMGSD: Multi-Modal Gaussian Shape Descriptors for Correspondence Matching in 1D and 2D Deformable Objects
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-03T06%3A54%3A59IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=MMGSD:%20Multi-Modal%20Gaussian%20Shape%20Descriptors%20for%20Correspondence%20Matching%20in%201D%20and%202D%20Deformable%20Objects&rft.au=Ganapathi,%20Aditya&rft.date=2020-10-08&rft_id=info:doi/10.48550/arxiv.2010.04339&rft_dat=%3Carxiv_GOX%3E2010_04339%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true