Parameter-free Online Test-time Adaptation

Training state-of-the-art vision models has become prohibitively expensive for researchers and practitioners. For the sake of accessibility and resource reuse, it is important to focus on adapting these models to a variety of down-stream scenarios. An interesting and practical paradigm is online tes...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online) S. 8334 - 8343
Hauptverfasser: Boudiaf, Malik, Mueller, Romain, Ayed, Ismail Ben, Bertinetto, Luca
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.06.2022
Schlagworte:
ISSN:1063-6919
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Abstract Training state-of-the-art vision models has become prohibitively expensive for researchers and practitioners. For the sake of accessibility and resource reuse, it is important to focus on adapting these models to a variety of down-stream scenarios. An interesting and practical paradigm is online test-time adaptation, according to which training data is inaccessible, no labelled data from the test distribution is available, and adaptation can only happen at test time and on a handful of samples. In this paper, we investigate how test-time adaptation methods fare for a number of pre-trained models on a variety of real-world scenarios, significantly extending the way they have been originally evaluated. We show that they perform well only in narrowly-defined experimental setups and sometimes fail catastrophically when their hyperparameters are not selected for the same scenario in which they are being tested. Motivated by the inherent uncertainty around the conditions that will ultimately be encountered at test time, we propose a particularly "conservative" approach, which addresses the problem with a Laplacian Adjusted Maximum-likelihood Estimation (LAME) objective. By adapting the model's output (not its parameters), and solving our objective with an efficient concave-convex procedure, our approach exhibits a much higher average accuracy across scenarios than existing methods, while being notably faster and have a much lower memory footprint. The code is available at https://github.com/fiveai/LAME.
AbstractList Training state-of-the-art vision models has become prohibitively expensive for researchers and practitioners. For the sake of accessibility and resource reuse, it is important to focus on adapting these models to a variety of down-stream scenarios. An interesting and practical paradigm is online test-time adaptation, according to which training data is inaccessible, no labelled data from the test distribution is available, and adaptation can only happen at test time and on a handful of samples. In this paper, we investigate how test-time adaptation methods fare for a number of pre-trained models on a variety of real-world scenarios, significantly extending the way they have been originally evaluated. We show that they perform well only in narrowly-defined experimental setups and sometimes fail catastrophically when their hyperparameters are not selected for the same scenario in which they are being tested. Motivated by the inherent uncertainty around the conditions that will ultimately be encountered at test time, we propose a particularly "conservative" approach, which addresses the problem with a Laplacian Adjusted Maximum-likelihood Estimation (LAME) objective. By adapting the model's output (not its parameters), and solving our objective with an efficient concave-convex procedure, our approach exhibits a much higher average accuracy across scenarios than existing methods, while being notably faster and have a much lower memory footprint. The code is available at https://github.com/fiveai/LAME.
Author Mueller, Romain
Ayed, Ismail Ben
Bertinetto, Luca
Boudiaf, Malik
Author_xml – sequence: 1
  givenname: Malik
  surname: Boudiaf
  fullname: Boudiaf, Malik
  organization: ÉTS Montreal
– sequence: 2
  givenname: Romain
  surname: Mueller
  fullname: Mueller, Romain
  organization: FiveAI
– sequence: 3
  givenname: Ismail Ben
  surname: Ayed
  fullname: Ayed, Ismail Ben
  organization: ÉTS Montreal
– sequence: 4
  givenname: Luca
  surname: Bertinetto
  fullname: Bertinetto, Luca
  organization: FiveAI
BookMark eNotzE1Lw0AQANBVFGxrf4EechY2zuz3HkuxKhRapHotk2QWVpq0JHvx33vQ07u9ubgZzgML8YhQI0J8Xn_tP6xyIdQKlKoBArorMUfnrHHROH0tZghOSxcx3onlNH0DgFaILoaZeNrTSD0XHmUamavdcMoDVweeiiy552rV0aVQyefhXtwmOk28_HchPjcvh_Wb3O5e39errcwKdJFo28SWomkarW3yJkZORJ6TTwRNaH2XsLPGBkBgJuBgQ6e5JUohtqgX4uHvzcx8vIy5p_HnGIOPXhn9C08DROQ
CODEN IEEPAD
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/CVPR52688.2022.00816
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP) 1998-present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Applied Sciences
EISBN 1665469463
9781665469463
EISSN 1063-6919
EndPage 8343
ExternalDocumentID 9879724
Genre orig-research
GroupedDBID 6IE
6IH
6IL
6IN
AAWTH
ABLEC
ADZIZ
ALMA_UNASSIGNED_HOLDINGS
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CBEJK
CHZPO
IEGSK
IJVOP
OCL
RIE
RIL
RIO
ID FETCH-LOGICAL-i203t-15cfe5a94bb335f7499efaa7ef7fa0b8c7df1d5458010eea0e858d3ecaaf89c13
IEDL.DBID RIE
ISICitedReferencesCount 96
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000870759101038&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
IngestDate Wed Aug 27 02:15:09 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-i203t-15cfe5a94bb335f7499efaa7ef7fa0b8c7df1d5458010eea0e858d3ecaaf89c13
PageCount 10
ParticipantIDs ieee_primary_9879724
PublicationCentury 2000
PublicationDate 2022-June
PublicationDateYYYYMMDD 2022-06-01
PublicationDate_xml – month: 06
  year: 2022
  text: 2022-June
PublicationDecade 2020
PublicationTitle Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online)
PublicationTitleAbbrev CVPR
PublicationYear 2022
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0003211698
Score 2.613043
Snippet Training state-of-the-art vision models has become prohibitively expensive for researchers and practitioners. For the sake of accessibility and resource reuse,...
SourceID ieee
SourceType Publisher
StartPage 8334
SubjectTerms Adaptation models
Laplace equations
Maximum likelihood estimation
Memory management
Protocols
Self-& semi-& meta- Efficient learning and inferences; Transfer/low-shot/long-tail learning
Training
Uncertainty
Title Parameter-free Online Test-time Adaptation
URI https://ieeexplore.ieee.org/document/9879724
WOSCitedRecordID wos000870759101038&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3PS8MwFH5sw4OnqZv4mxw8iXFtky7JUYbDg4wiU3YbafICu3Sj6_z7TdpSPXjxFgIhJOHl5ct77_sA7q1hyvp3KDWRVJQj097mFNLEBe1o72Bcw67_JhYLuVqprAePXS0MItbJZ_gUmnUs327NIXyVTTw-ViLhfegLMW1qtbr_FOaRzFTJtjoujtRk9pm9BzKTkMCVBFpOGUTNf2mo1C5kPvzf5Ccw_qnFI1nnZU6hh8UZDNvHI2lNcz-Ch0yHPCu_TdSViKThECVLf-vTICBPnq3eNXH3MXzMX5azV9oKIdBNErGKxqlxmGrF85yx1AmPUtBpLdAJp6NcGmFdbEMIzKMrRB2hTKVlaLR2UpmYncOg2BZ4ASTChDOOzsM6zRnjyg_Jc4mxTVBqNJcwCktf7xqui3W76qu_u6_hOOxtkzp1A4OqPOAtHJmvarMv7-oD-gZ4-ZGF
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NSwMxEB1qFfRUtRW_3YMnMTabZN3kKMVSsZZFqvRWsskEemnLtvX3m-wu1YMXbyEQQhImk5eZeQ_g1hqurH-HEkOlIgK59jankDAXtKO9g3EVu_4wHY3kZKKyBtxva2EQsUw-w4fQLGP5dmE24aus6_GxSpnYgd1ECEaraq3tjwr3WOZRybo-Lqaq2_vM3gOdSUjhYoGYUwZZ818qKqUT6bf-N_0hdH6q8aJs62eOoIHzY2jVz8eoNs5VG-4yHTKt_EYRVyBGFYtoNPb3PgkS8tGT1csq8t6Bj_7zuDcgtRQCmTHK1yROjMNEK5HnnCcu9TgFndYputRpmkuTWhfbEATz-ApRU5SJtByN1k4qE_MTaM4XczyFiCITXKDzwE4LzoXyQ_JcYmwZSo3mDNph6dNlxXYxrVd9_nf3DewPxm_D6fBl9HoBB2Gfq0SqS2iuiw1ewZ75Ws9WxXV5WN88qpTM
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=proceeding&rft.title=Proceedings+%28IEEE+Computer+Society+Conference+on+Computer+Vision+and+Pattern+Recognition.+Online%29&rft.atitle=Parameter-free+Online+Test-time+Adaptation&rft.au=Boudiaf%2C+Malik&rft.au=Mueller%2C+Romain&rft.au=Ayed%2C+Ismail+Ben&rft.au=Bertinetto%2C+Luca&rft.date=2022-06-01&rft.pub=IEEE&rft.eissn=1063-6919&rft.spage=8334&rft.epage=8343&rft_id=info:doi/10.1109%2FCVPR52688.2022.00816&rft.externalDocID=9879724