Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation

Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classifier to adapt to the target domain and do not properly handle the trade off between the source domain and the target domain. In this work, instead...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Ye, Shaokai, Wu, Kailu, Zhou, Mu, Yang, Yunfei, Tan, Sia huat, Xu, Kaidi, Song, Jiebo, Bao, Chenglong, Ma, Kaisheng
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 Ye, Shaokai
Wu, Kailu
Zhou, Mu
Yang, Yunfei
Tan, Sia huat
Xu, Kaidi
Song, Jiebo
Bao, Chenglong
Ma, Kaisheng
description Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classifier to adapt to the target domain and do not properly handle the trade off between the source domain and the target domain. In this work, instead of training a classifier to adapt to the target domain, we use a separable component called data calibrator to help the fixed source classifier recover discrimination power in the target domain, while preserving the source domain's performance. When the difference between two domains is small, the source classifier's representation is sufficient to perform well in the target domain and outperforms GAN-based methods in digits. Otherwise, the proposed method can leverage synthetic images generated by GANs to boost performance and achieve state-of-the-art performance in digits datasets and driving scene semantic segmentation. Our method empirically reveals that certain intriguing hints, which can be mitigated by adversarial attack to domain discriminators, are one of the sources for performance degradation under the domain shift.
doi_str_mv 10.48550/arxiv.1911.12796
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1911_12796</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1911_12796</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-e8030501518473498bb2c6e64d547f0f22edb1eb630ebcfb314f80ea6f16bcf43</originalsourceid><addsrcrecordid>eNotz0tOwzAUhWFPGKDCAhjhDSTY8SMOsyo8pUgI0Y6j6-a6WErsyDEFdg8tjH6dyZE-Qq44K6VRit1A-vKHkjecl7yqG31OXju_f8_FJx5DWxi9TZBjuqVA33CGBHZE2sZpjgFDpi4mug3Lx4zp4Bcc6F2cwAe6HmDOkH0MF-TMwbjg5X9XZPNwv2mfiu7l8blddwXoWhdomGCKccWNrIVsjLXVTqOWg5K1Y66qcLAcrRYM7c5ZwaUzDEE7rn-3FCty_Xd7MvVz8hOk7_5o60828QOay0pF</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation</title><source>arXiv.org</source><creator>Ye, Shaokai ; Wu, Kailu ; Zhou, Mu ; Yang, Yunfei ; Tan, Sia huat ; Xu, Kaidi ; Song, Jiebo ; Bao, Chenglong ; Ma, Kaisheng</creator><creatorcontrib>Ye, Shaokai ; Wu, Kailu ; Zhou, Mu ; Yang, Yunfei ; Tan, Sia huat ; Xu, Kaidi ; Song, Jiebo ; Bao, Chenglong ; Ma, Kaisheng</creatorcontrib><description>Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classifier to adapt to the target domain and do not properly handle the trade off between the source domain and the target domain. In this work, instead of training a classifier to adapt to the target domain, we use a separable component called data calibrator to help the fixed source classifier recover discrimination power in the target domain, while preserving the source domain's performance. When the difference between two domains is small, the source classifier's representation is sufficient to perform well in the target domain and outperforms GAN-based methods in digits. Otherwise, the proposed method can leverage synthetic images generated by GANs to boost performance and achieve state-of-the-art performance in digits datasets and driving scene semantic segmentation. Our method empirically reveals that certain intriguing hints, which can be mitigated by adversarial attack to domain discriminators, are one of the sources for performance degradation under the domain shift.</description><identifier>DOI: 10.48550/arxiv.1911.12796</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Learning</subject><creationdate>2019-11</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,781,886</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/1911.12796$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1911.12796$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Ye, Shaokai</creatorcontrib><creatorcontrib>Wu, Kailu</creatorcontrib><creatorcontrib>Zhou, Mu</creatorcontrib><creatorcontrib>Yang, Yunfei</creatorcontrib><creatorcontrib>Tan, Sia huat</creatorcontrib><creatorcontrib>Xu, Kaidi</creatorcontrib><creatorcontrib>Song, Jiebo</creatorcontrib><creatorcontrib>Bao, Chenglong</creatorcontrib><creatorcontrib>Ma, Kaisheng</creatorcontrib><title>Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation</title><description>Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classifier to adapt to the target domain and do not properly handle the trade off between the source domain and the target domain. In this work, instead of training a classifier to adapt to the target domain, we use a separable component called data calibrator to help the fixed source classifier recover discrimination power in the target domain, while preserving the source domain's performance. When the difference between two domains is small, the source classifier's representation is sufficient to perform well in the target domain and outperforms GAN-based methods in digits. Otherwise, the proposed method can leverage synthetic images generated by GANs to boost performance and achieve state-of-the-art performance in digits datasets and driving scene semantic segmentation. Our method empirically reveals that certain intriguing hints, which can be mitigated by adversarial attack to domain discriminators, are one of the sources for performance degradation under the domain shift.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2019</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz0tOwzAUhWFPGKDCAhjhDSTY8SMOsyo8pUgI0Y6j6-a6WErsyDEFdg8tjH6dyZE-Qq44K6VRit1A-vKHkjecl7yqG31OXju_f8_FJx5DWxi9TZBjuqVA33CGBHZE2sZpjgFDpi4mug3Lx4zp4Bcc6F2cwAe6HmDOkH0MF-TMwbjg5X9XZPNwv2mfiu7l8blddwXoWhdomGCKccWNrIVsjLXVTqOWg5K1Y66qcLAcrRYM7c5ZwaUzDEE7rn-3FCty_Xd7MvVz8hOk7_5o60828QOay0pF</recordid><startdate>20191128</startdate><enddate>20191128</enddate><creator>Ye, Shaokai</creator><creator>Wu, Kailu</creator><creator>Zhou, Mu</creator><creator>Yang, Yunfei</creator><creator>Tan, Sia huat</creator><creator>Xu, Kaidi</creator><creator>Song, Jiebo</creator><creator>Bao, Chenglong</creator><creator>Ma, Kaisheng</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20191128</creationdate><title>Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation</title><author>Ye, Shaokai ; Wu, Kailu ; Zhou, Mu ; Yang, Yunfei ; Tan, Sia huat ; Xu, Kaidi ; Song, Jiebo ; Bao, Chenglong ; Ma, Kaisheng</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-e8030501518473498bb2c6e64d547f0f22edb1eb630ebcfb314f80ea6f16bcf43</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2019</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Ye, Shaokai</creatorcontrib><creatorcontrib>Wu, Kailu</creatorcontrib><creatorcontrib>Zhou, Mu</creatorcontrib><creatorcontrib>Yang, Yunfei</creatorcontrib><creatorcontrib>Tan, Sia huat</creatorcontrib><creatorcontrib>Xu, Kaidi</creatorcontrib><creatorcontrib>Song, Jiebo</creatorcontrib><creatorcontrib>Bao, Chenglong</creatorcontrib><creatorcontrib>Ma, Kaisheng</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Ye, Shaokai</au><au>Wu, Kailu</au><au>Zhou, Mu</au><au>Yang, Yunfei</au><au>Tan, Sia huat</au><au>Xu, Kaidi</au><au>Song, Jiebo</au><au>Bao, Chenglong</au><au>Ma, Kaisheng</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation</atitle><date>2019-11-28</date><risdate>2019</risdate><abstract>Existing domain adaptation methods aim at learning features that can be generalized among domains. These methods commonly require to update source classifier to adapt to the target domain and do not properly handle the trade off between the source domain and the target domain. In this work, instead of training a classifier to adapt to the target domain, we use a separable component called data calibrator to help the fixed source classifier recover discrimination power in the target domain, while preserving the source domain's performance. When the difference between two domains is small, the source classifier's representation is sufficient to perform well in the target domain and outperforms GAN-based methods in digits. Otherwise, the proposed method can leverage synthetic images generated by GANs to boost performance and achieve state-of-the-art performance in digits datasets and driving scene semantic segmentation. Our method empirically reveals that certain intriguing hints, which can be mitigated by adversarial attack to domain discriminators, are one of the sources for performance degradation under the domain shift.</abstract><doi>10.48550/arxiv.1911.12796</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1911.12796
ispartof
issn
language eng
recordid cdi_arxiv_primary_1911_12796
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
Computer Science - Learning
title Light-weight Calibrator: a Separable Component for Unsupervised Domain Adaptation
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-17T01%3A15%3A24IST&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=Light-weight%20Calibrator:%20a%20Separable%20Component%20for%20Unsupervised%20Domain%20Adaptation&rft.au=Ye,%20Shaokai&rft.date=2019-11-28&rft_id=info:doi/10.48550/arxiv.1911.12796&rft_dat=%3Carxiv_GOX%3E1911_12796%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