Robot Instance Segmentation with Few Annotations for Grasping

The ability of robots to manipulate objects relies heavily on their aptitude for visual perception. In domains characterized by cluttered scenes and high object variability, most methods call for vast labeled datasets, laboriously hand-annotated, with the aim of training capable models. Once deploye...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kimhi, Moshe, Vainshtein, David, Baskin, Chaim, Di Castro, Dotan
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 Kimhi, Moshe
Vainshtein, David
Baskin, Chaim
Di Castro, Dotan
description The ability of robots to manipulate objects relies heavily on their aptitude for visual perception. In domains characterized by cluttered scenes and high object variability, most methods call for vast labeled datasets, laboriously hand-annotated, with the aim of training capable models. Once deployed, the challenge of generalizing to unfamiliar objects implies that the model must evolve alongside its domain. To address this, we propose a novel framework that combines Semi-Supervised Learning (SSL) with Learning Through Interaction (LTI), allowing a model to learn by observing scene alterations and leverage visual consistency despite temporal gaps without requiring curated data of interaction sequences. As a result, our approach exploits partially annotated data through self-supervision and incorporates temporal context using pseudo-sequences generated from unlabeled still images. We validate our method on two common benchmarks, ARMBench mix-object-tote and OCID, where it achieves state-of-the-art performance. Notably, on ARMBench, we attain an $\text{AP}_{50}$ of $86.37$, almost a $20\%$ improvement over existing work, and obtain remarkable results in scenarios with extremely low annotation, achieving an $\text{AP}_{50}$ score of $84.89$ with just $1 \%$ of annotated data compared to $72$ presented in ARMBench on the fully annotated counterpart.
doi_str_mv 10.48550/arxiv.2407.01302
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2407_01302</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2407_01302</sourcerecordid><originalsourceid>FETCH-arxiv_primary_2407_013023</originalsourceid><addsrcrecordid>eNpjYJA0NNAzsTA1NdBPLKrILNMzMjEw1zMwNDYw4mSwDcpPyi9R8MwrLknMS05VCE5Nz03NK0ksyczPUyjPLMlQcEstV3DMy8uHiBUrpOUXKbgXJRYXZOal8zCwpiXmFKfyQmluBnk31xBnD12wRfEFRZm5iUWV8SAL48EWGhNWAQBGlTWh</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Robot Instance Segmentation with Few Annotations for Grasping</title><source>arXiv.org</source><creator>Kimhi, Moshe ; Vainshtein, David ; Baskin, Chaim ; Di Castro, Dotan</creator><creatorcontrib>Kimhi, Moshe ; Vainshtein, David ; Baskin, Chaim ; Di Castro, Dotan</creatorcontrib><description>The ability of robots to manipulate objects relies heavily on their aptitude for visual perception. In domains characterized by cluttered scenes and high object variability, most methods call for vast labeled datasets, laboriously hand-annotated, with the aim of training capable models. Once deployed, the challenge of generalizing to unfamiliar objects implies that the model must evolve alongside its domain. To address this, we propose a novel framework that combines Semi-Supervised Learning (SSL) with Learning Through Interaction (LTI), allowing a model to learn by observing scene alterations and leverage visual consistency despite temporal gaps without requiring curated data of interaction sequences. As a result, our approach exploits partially annotated data through self-supervision and incorporates temporal context using pseudo-sequences generated from unlabeled still images. We validate our method on two common benchmarks, ARMBench mix-object-tote and OCID, where it achieves state-of-the-art performance. Notably, on ARMBench, we attain an $\text{AP}_{50}$ of $86.37$, almost a $20\%$ improvement over existing work, and obtain remarkable results in scenarios with extremely low annotation, achieving an $\text{AP}_{50}$ score of $84.89$ with just $1 \%$ of annotated data compared to $72$ presented in ARMBench on the fully annotated counterpart.</description><identifier>DOI: 10.48550/arxiv.2407.01302</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Robotics</subject><creationdate>2024-07</creationdate><rights>http://creativecommons.org/licenses/by-nc-sa/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2407.01302$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2407.01302$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Kimhi, Moshe</creatorcontrib><creatorcontrib>Vainshtein, David</creatorcontrib><creatorcontrib>Baskin, Chaim</creatorcontrib><creatorcontrib>Di Castro, Dotan</creatorcontrib><title>Robot Instance Segmentation with Few Annotations for Grasping</title><description>The ability of robots to manipulate objects relies heavily on their aptitude for visual perception. In domains characterized by cluttered scenes and high object variability, most methods call for vast labeled datasets, laboriously hand-annotated, with the aim of training capable models. Once deployed, the challenge of generalizing to unfamiliar objects implies that the model must evolve alongside its domain. To address this, we propose a novel framework that combines Semi-Supervised Learning (SSL) with Learning Through Interaction (LTI), allowing a model to learn by observing scene alterations and leverage visual consistency despite temporal gaps without requiring curated data of interaction sequences. As a result, our approach exploits partially annotated data through self-supervision and incorporates temporal context using pseudo-sequences generated from unlabeled still images. We validate our method on two common benchmarks, ARMBench mix-object-tote and OCID, where it achieves state-of-the-art performance. Notably, on ARMBench, we attain an $\text{AP}_{50}$ of $86.37$, almost a $20\%$ improvement over existing work, and obtain remarkable results in scenarios with extremely low annotation, achieving an $\text{AP}_{50}$ score of $84.89$ with just $1 \%$ of annotated data compared to $72$ presented in ARMBench on the fully annotated counterpart.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Robotics</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNpjYJA0NNAzsTA1NdBPLKrILNMzMjEw1zMwNDYw4mSwDcpPyi9R8MwrLknMS05VCE5Nz03NK0ksyczPUyjPLMlQcEstV3DMy8uHiBUrpOUXKbgXJRYXZOal8zCwpiXmFKfyQmluBnk31xBnD12wRfEFRZm5iUWV8SAL48EWGhNWAQBGlTWh</recordid><startdate>20240701</startdate><enddate>20240701</enddate><creator>Kimhi, Moshe</creator><creator>Vainshtein, David</creator><creator>Baskin, Chaim</creator><creator>Di Castro, Dotan</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20240701</creationdate><title>Robot Instance Segmentation with Few Annotations for Grasping</title><author>Kimhi, Moshe ; Vainshtein, David ; Baskin, Chaim ; Di Castro, Dotan</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-arxiv_primary_2407_013023</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Robotics</topic><toplevel>online_resources</toplevel><creatorcontrib>Kimhi, Moshe</creatorcontrib><creatorcontrib>Vainshtein, David</creatorcontrib><creatorcontrib>Baskin, Chaim</creatorcontrib><creatorcontrib>Di Castro, Dotan</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kimhi, Moshe</au><au>Vainshtein, David</au><au>Baskin, Chaim</au><au>Di Castro, Dotan</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Robot Instance Segmentation with Few Annotations for Grasping</atitle><date>2024-07-01</date><risdate>2024</risdate><abstract>The ability of robots to manipulate objects relies heavily on their aptitude for visual perception. In domains characterized by cluttered scenes and high object variability, most methods call for vast labeled datasets, laboriously hand-annotated, with the aim of training capable models. Once deployed, the challenge of generalizing to unfamiliar objects implies that the model must evolve alongside its domain. To address this, we propose a novel framework that combines Semi-Supervised Learning (SSL) with Learning Through Interaction (LTI), allowing a model to learn by observing scene alterations and leverage visual consistency despite temporal gaps without requiring curated data of interaction sequences. As a result, our approach exploits partially annotated data through self-supervision and incorporates temporal context using pseudo-sequences generated from unlabeled still images. We validate our method on two common benchmarks, ARMBench mix-object-tote and OCID, where it achieves state-of-the-art performance. Notably, on ARMBench, we attain an $\text{AP}_{50}$ of $86.37$, almost a $20\%$ improvement over existing work, and obtain remarkable results in scenarios with extremely low annotation, achieving an $\text{AP}_{50}$ score of $84.89$ with just $1 \%$ of annotated data compared to $72$ presented in ARMBench on the fully annotated counterpart.</abstract><doi>10.48550/arxiv.2407.01302</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2407.01302
ispartof
issn
language eng
recordid cdi_arxiv_primary_2407_01302
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Computer Vision and Pattern Recognition
Computer Science - Robotics
title Robot Instance Segmentation with Few Annotations for Grasping
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-30T13%3A28%3A27IST&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=Robot%20Instance%20Segmentation%20with%20Few%20Annotations%20for%20Grasping&rft.au=Kimhi,%20Moshe&rft.date=2024-07-01&rft_id=info:doi/10.48550/arxiv.2407.01302&rft_dat=%3Carxiv_GOX%3E2407_01302%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