Learning object class detectors from weakly annotated video

Object detectors are typically trained on a large set of still images annotated by bounding-boxes. This paper introduces an approach for learning object detectors from real-world web videos known only to contain objects of a target class. We propose a fully automatic pipeline that localizes objects...

Full description

Saved in:
Bibliographic Details
Main Authors: Prest, Alessandro, Leistner, C., Civera, J., Schmid, C., Ferrari, V.
Format: Conference Proceeding
Language:English
Subjects:
Detectors
Electron tubes
Hidden Markov models
Image segmentation
Motion segmentation
Tracking
Training
Citations: Items that cite this one
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by cdi_FETCH-LOGICAL-c300t-b6a2511271f169834198738de66941fa0a4ef59597e5448fe8b9962311dc252d3
cites
container_end_page 3289
container_issue
container_start_page 3282
container_title
container_volume
creator Prest, Alessandro
Leistner, C.
Civera, J.
Schmid, C.
Ferrari, V.
description Object detectors are typically trained on a large set of still images annotated by bounding-boxes. This paper introduces an approach for learning object detectors from real-world web videos known only to contain objects of a target class. We propose a fully automatic pipeline that localizes objects in a set of videos of the class and learns a detector for it. The approach extracts candidate spatio-temporal tubes based on motion segmentation and then selects one tube per video jointly over all videos. To compare to the state of the art, we test our detector on still images, i.e., Pascal VOC 2007. We observe that frames extracted from web videos can differ significantly in terms of quality to still images taken by a good camera. Thus, we formulate the learning from videos as a domain adaptation task. We show that training from a combination of weakly annotated videos and fully annotated still images using domain adaptation improves the performance of a detector trained from still images alone.
doi_str_mv 10.1109/CVPR.2012.6248065
format conference_proceeding
fullrecord <record><control><sourceid>ieee_6IE</sourceid><recordid>TN_cdi_ieee_primary_6248065</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>6248065</ieee_id><sourcerecordid>6248065</sourcerecordid><originalsourceid>FETCH-LOGICAL-c300t-b6a2511271f169834198738de66941fa0a4ef59597e5448fe8b9962311dc252d3</originalsourceid><addsrcrecordid>eNo1j81KxDAUhSMqOI59AHGTF2jNvflpgispjgoFRdTtkDY30nGmlaYo8_YWHM_m8MHHgcPYJYgCQLjr6v35pUABWBhUVhh9xM5BmVICosVjlrnS_rNRJ2wBwsjcOHBnLEtpI-bMhnC4YDc1-bHv-g8-NBtqJ95ufUo80DTDMCYex2HHf8h_bvfc9_0w-YkC_-4CDRfsNPptouzQS_a2unutHvL66f6xuq3zVgox5Y3xqAGwhAjGWanA2VLaQMY4BdELryhqp11JWikbyTbOGZQAoUWNQS7Z1d9uR0Trr7Hb-XG_PnyXv_LmSWk</addsrcrecordid><sourcetype>Publisher</sourcetype><iscdi>true</iscdi><recordtype>conference_proceeding</recordtype></control><display><type>conference_proceeding</type><title>Learning object class detectors from weakly annotated video</title><source>IEEE Electronic Library (IEL) Conference Proceedings</source><creator>Prest, Alessandro ; Leistner, C. ; Civera, J. ; Schmid, C. ; Ferrari, V.</creator><creatorcontrib>Prest, Alessandro ; Leistner, C. ; Civera, J. ; Schmid, C. ; Ferrari, V.</creatorcontrib><description>Object detectors are typically trained on a large set of still images annotated by bounding-boxes. This paper introduces an approach for learning object detectors from real-world web videos known only to contain objects of a target class. We propose a fully automatic pipeline that localizes objects in a set of videos of the class and learns a detector for it. The approach extracts candidate spatio-temporal tubes based on motion segmentation and then selects one tube per video jointly over all videos. To compare to the state of the art, we test our detector on still images, i.e., Pascal VOC 2007. We observe that frames extracted from web videos can differ significantly in terms of quality to still images taken by a good camera. Thus, we formulate the learning from videos as a domain adaptation task. We show that training from a combination of weakly annotated videos and fully annotated still images using domain adaptation improves the performance of a detector trained from still images alone.</description><identifier>ISSN: 1063-6919</identifier><identifier>ISBN: 9781467312264</identifier><identifier>ISBN: 1467312266</identifier><identifier>EISBN: 1467312282</identifier><identifier>EISBN: 1467312274</identifier><identifier>EISBN: 9781467312271</identifier><identifier>EISBN: 9781467312288</identifier><identifier>DOI: 10.1109/CVPR.2012.6248065</identifier><language>eng</language><publisher>IEEE</publisher><subject>Detectors ; Electron tubes ; Hidden Markov models ; Image segmentation ; Motion segmentation ; Tracking ; Training</subject><ispartof>2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012, p.3282-3289</ispartof><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c300t-b6a2511271f169834198738de66941fa0a4ef59597e5448fe8b9962311dc252d3</citedby></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/6248065$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>310,311,786,790,795,796,2071,27958,55271</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/6248065$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Prest, Alessandro</creatorcontrib><creatorcontrib>Leistner, C.</creatorcontrib><creatorcontrib>Civera, J.</creatorcontrib><creatorcontrib>Schmid, C.</creatorcontrib><creatorcontrib>Ferrari, V.</creatorcontrib><title>Learning object class detectors from weakly annotated video</title><title>2012 IEEE Conference on Computer Vision and Pattern Recognition</title><addtitle>CVPR</addtitle><description>Object detectors are typically trained on a large set of still images annotated by bounding-boxes. This paper introduces an approach for learning object detectors from real-world web videos known only to contain objects of a target class. We propose a fully automatic pipeline that localizes objects in a set of videos of the class and learns a detector for it. The approach extracts candidate spatio-temporal tubes based on motion segmentation and then selects one tube per video jointly over all videos. To compare to the state of the art, we test our detector on still images, i.e., Pascal VOC 2007. We observe that frames extracted from web videos can differ significantly in terms of quality to still images taken by a good camera. Thus, we formulate the learning from videos as a domain adaptation task. We show that training from a combination of weakly annotated videos and fully annotated still images using domain adaptation improves the performance of a detector trained from still images alone.</description><subject>Detectors</subject><subject>Electron tubes</subject><subject>Hidden Markov models</subject><subject>Image segmentation</subject><subject>Motion segmentation</subject><subject>Tracking</subject><subject>Training</subject><issn>1063-6919</issn><isbn>9781467312264</isbn><isbn>1467312266</isbn><isbn>1467312282</isbn><isbn>1467312274</isbn><isbn>9781467312271</isbn><isbn>9781467312288</isbn><fulltext>true</fulltext><rsrctype>conference_proceeding</rsrctype><creationdate>2012</creationdate><recordtype>conference_proceeding</recordtype><sourceid>6IE</sourceid><recordid>eNo1j81KxDAUhSMqOI59AHGTF2jNvflpgispjgoFRdTtkDY30nGmlaYo8_YWHM_m8MHHgcPYJYgCQLjr6v35pUABWBhUVhh9xM5BmVICosVjlrnS_rNRJ2wBwsjcOHBnLEtpI-bMhnC4YDc1-bHv-g8-NBtqJ95ufUo80DTDMCYex2HHf8h_bvfc9_0w-YkC_-4CDRfsNPptouzQS_a2unutHvL66f6xuq3zVgox5Y3xqAGwhAjGWanA2VLaQMY4BdELryhqp11JWikbyTbOGZQAoUWNQS7Z1d9uR0Trr7Hb-XG_PnyXv_LmSWk</recordid><startdate>201206</startdate><enddate>201206</enddate><creator>Prest, Alessandro</creator><creator>Leistner, C.</creator><creator>Civera, J.</creator><creator>Schmid, C.</creator><creator>Ferrari, V.</creator><general>IEEE</general><scope>6IE</scope><scope>6IH</scope><scope>CBEJK</scope><scope>RIE</scope><scope>RIO</scope></search><sort><creationdate>201206</creationdate><title>Learning object class detectors from weakly annotated video</title><author>Prest, Alessandro ; Leistner, C. ; Civera, J. ; Schmid, C. ; Ferrari, V.</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c300t-b6a2511271f169834198738de66941fa0a4ef59597e5448fe8b9962311dc252d3</frbrgroupid><rsrctype>conference_proceedings</rsrctype><prefilter>conference_proceedings</prefilter><language>eng</language><creationdate>2012</creationdate><topic>Detectors</topic><topic>Electron tubes</topic><topic>Hidden Markov models</topic><topic>Image segmentation</topic><topic>Motion segmentation</topic><topic>Tracking</topic><topic>Training</topic><toplevel>online_resources</toplevel><creatorcontrib>Prest, Alessandro</creatorcontrib><creatorcontrib>Leistner, C.</creatorcontrib><creatorcontrib>Civera, J.</creatorcontrib><creatorcontrib>Schmid, C.</creatorcontrib><creatorcontrib>Ferrari, V.</creatorcontrib><collection>IEEE Electronic Library (IEL) Conference Proceedings</collection><collection>IEEE Proceedings Order Plan (POP) 1998-present by volume</collection><collection>IEEE Xplore All Conference Proceedings</collection><collection>IEEE Electronic Library (IEL)</collection><collection>IEEE Proceedings Order Plans (POP) 1998-present</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Prest, Alessandro</au><au>Leistner, C.</au><au>Civera, J.</au><au>Schmid, C.</au><au>Ferrari, V.</au><format>book</format><genre>proceeding</genre><ristype>CONF</ristype><atitle>Learning object class detectors from weakly annotated video</atitle><btitle>2012 IEEE Conference on Computer Vision and Pattern Recognition</btitle><stitle>CVPR</stitle><date>2012-06</date><risdate>2012</risdate><spage>3282</spage><epage>3289</epage><pages>3282-3289</pages><issn>1063-6919</issn><isbn>9781467312264</isbn><isbn>1467312266</isbn><eisbn>1467312282</eisbn><eisbn>1467312274</eisbn><eisbn>9781467312271</eisbn><eisbn>9781467312288</eisbn><abstract>Object detectors are typically trained on a large set of still images annotated by bounding-boxes. This paper introduces an approach for learning object detectors from real-world web videos known only to contain objects of a target class. We propose a fully automatic pipeline that localizes objects in a set of videos of the class and learns a detector for it. The approach extracts candidate spatio-temporal tubes based on motion segmentation and then selects one tube per video jointly over all videos. To compare to the state of the art, we test our detector on still images, i.e., Pascal VOC 2007. We observe that frames extracted from web videos can differ significantly in terms of quality to still images taken by a good camera. Thus, we formulate the learning from videos as a domain adaptation task. We show that training from a combination of weakly annotated videos and fully annotated still images using domain adaptation improves the performance of a detector trained from still images alone.</abstract><pub>IEEE</pub><doi>10.1109/CVPR.2012.6248065</doi><tpages>8</tpages><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier ISSN: 1063-6919
ispartof 2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012, p.3282-3289
issn 1063-6919
language eng
recordid cdi_ieee_primary_6248065
source IEEE Electronic Library (IEL) Conference Proceedings
subjects Detectors
Electron tubes
Hidden Markov models
Image segmentation
Motion segmentation
Tracking
Training
title Learning object class detectors from weakly annotated video
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-09-21T17%3A32%3A36IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-ieee_6IE&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=proceeding&rft.atitle=Learning%20object%20class%20detectors%20from%20weakly%20annotated%20video&rft.btitle=2012%20IEEE%20Conference%20on%20Computer%20Vision%20and%20Pattern%20Recognition&rft.au=Prest,%20Alessandro&rft.date=2012-06&rft.spage=3282&rft.epage=3289&rft.pages=3282-3289&rft.issn=1063-6919&rft.isbn=9781467312264&rft.isbn_list=1467312266&rft_id=info:doi/10.1109/CVPR.2012.6248065&rft.eisbn=1467312282&rft.eisbn_list=1467312274&rft.eisbn_list=9781467312271&rft.eisbn_list=9781467312288&rft_dat=%3Cieee_6IE%3E6248065%3C/ieee_6IE%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c300t-b6a2511271f169834198738de66941fa0a4ef59597e5448fe8b9962311dc252d3%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=6248065&rfr_iscdi=true