Adjustment learning and relevant component analysis

Noam Shental, Tomer Hertz, Daphna Weinshall, Misha Pavel

Research output: Chapter in Book/Report/Conference proceedingConference contribution

126 Citations (Scopus)

Abstract

We propose a new learning approach for image retrieval, which we call adjustment learning, and demonstrate its use for face recognition and color matching. Our approach is motivated by a frequently encountered problem, namely, that variability in the original data representation which is not relevant to the task may interfere with retrieval and make it very difficult. Our key observation is that in real applications of image retrieval, data sometimes comes in small chunks - small subsets of images that come from the same (but unknown) class. This is the case, for example, when a query is presented via a short video clip. We call these groups chunklets, and we call the paradigm which uses chunklets for unsupervised learning adjustment learning. Within this paradigm we propose a linear scheme, which we call Relevant Component Analysis; this scheme uses the information in such chunklets to reduce irrelevant variability in the data while amplifying relevant variability. We provide results using our method on two problems: face recognition (using a database publicly available on the web), and visual surveillance (using our own data). In the latter application chunklets are obtained automatically from the data without the need of supervision.

Original languageEnglish (US)
Title of host publicationComputer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings
PublisherSpringer Verlag
Pages776-790
Number of pages15
Volume2353
ISBN (Print)9783540437482
StatePublished - 2002
Event7th European Conference on Computer Vision, ECCV 2002 - Copenhagen, Denmark
Duration: May 28 2002May 31 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2353
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other7th European Conference on Computer Vision, ECCV 2002
CountryDenmark
CityCopenhagen
Period5/28/025/31/02

Fingerprint

Image retrieval
Face recognition
Adjustment
Color matching
Unsupervised learning
Image Retrieval
Face Recognition
Paradigm
Visual Surveillance
Unsupervised Learning
Retrieval
Learning
Query
Unknown
Subset
Demonstrate

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Shental, N., Hertz, T., Weinshall, D., & Pavel, M. (2002). Adjustment learning and relevant component analysis. In Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings (Vol. 2353, pp. 776-790). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2353). Springer Verlag.

Adjustment learning and relevant component analysis. / Shental, Noam; Hertz, Tomer; Weinshall, Daphna; Pavel, Misha.

Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings. Vol. 2353 Springer Verlag, 2002. p. 776-790 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2353).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Shental, N, Hertz, T, Weinshall, D & Pavel, M 2002, Adjustment learning and relevant component analysis. in Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings. vol. 2353, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 2353, Springer Verlag, pp. 776-790, 7th European Conference on Computer Vision, ECCV 2002, Copenhagen, Denmark, 5/28/02.
Shental N, Hertz T, Weinshall D, Pavel M. Adjustment learning and relevant component analysis. In Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings. Vol. 2353. Springer Verlag. 2002. p. 776-790. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Shental, Noam ; Hertz, Tomer ; Weinshall, Daphna ; Pavel, Misha. / Adjustment learning and relevant component analysis. Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings. Vol. 2353 Springer Verlag, 2002. pp. 776-790 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{50587567a57d443d86cddc8b29fada31,
title = "Adjustment learning and relevant component analysis",
abstract = "We propose a new learning approach for image retrieval, which we call adjustment learning, and demonstrate its use for face recognition and color matching. Our approach is motivated by a frequently encountered problem, namely, that variability in the original data representation which is not relevant to the task may interfere with retrieval and make it very difficult. Our key observation is that in real applications of image retrieval, data sometimes comes in small chunks - small subsets of images that come from the same (but unknown) class. This is the case, for example, when a query is presented via a short video clip. We call these groups chunklets, and we call the paradigm which uses chunklets for unsupervised learning adjustment learning. Within this paradigm we propose a linear scheme, which we call Relevant Component Analysis; this scheme uses the information in such chunklets to reduce irrelevant variability in the data while amplifying relevant variability. We provide results using our method on two problems: face recognition (using a database publicly available on the web), and visual surveillance (using our own data). In the latter application chunklets are obtained automatically from the data without the need of supervision.",
author = "Noam Shental and Tomer Hertz and Daphna Weinshall and Misha Pavel",
year = "2002",
language = "English (US)",
isbn = "9783540437482",
volume = "2353",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "776--790",
booktitle = "Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings",

}

TY - GEN

T1 - Adjustment learning and relevant component analysis

AU - Shental, Noam

AU - Hertz, Tomer

AU - Weinshall, Daphna

AU - Pavel, Misha

PY - 2002

Y1 - 2002

N2 - We propose a new learning approach for image retrieval, which we call adjustment learning, and demonstrate its use for face recognition and color matching. Our approach is motivated by a frequently encountered problem, namely, that variability in the original data representation which is not relevant to the task may interfere with retrieval and make it very difficult. Our key observation is that in real applications of image retrieval, data sometimes comes in small chunks - small subsets of images that come from the same (but unknown) class. This is the case, for example, when a query is presented via a short video clip. We call these groups chunklets, and we call the paradigm which uses chunklets for unsupervised learning adjustment learning. Within this paradigm we propose a linear scheme, which we call Relevant Component Analysis; this scheme uses the information in such chunklets to reduce irrelevant variability in the data while amplifying relevant variability. We provide results using our method on two problems: face recognition (using a database publicly available on the web), and visual surveillance (using our own data). In the latter application chunklets are obtained automatically from the data without the need of supervision.

AB - We propose a new learning approach for image retrieval, which we call adjustment learning, and demonstrate its use for face recognition and color matching. Our approach is motivated by a frequently encountered problem, namely, that variability in the original data representation which is not relevant to the task may interfere with retrieval and make it very difficult. Our key observation is that in real applications of image retrieval, data sometimes comes in small chunks - small subsets of images that come from the same (but unknown) class. This is the case, for example, when a query is presented via a short video clip. We call these groups chunklets, and we call the paradigm which uses chunklets for unsupervised learning adjustment learning. Within this paradigm we propose a linear scheme, which we call Relevant Component Analysis; this scheme uses the information in such chunklets to reduce irrelevant variability in the data while amplifying relevant variability. We provide results using our method on two problems: face recognition (using a database publicly available on the web), and visual surveillance (using our own data). In the latter application chunklets are obtained automatically from the data without the need of supervision.

UR - http://www.scopus.com/inward/record.url?scp=84937544784&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937544784&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9783540437482

VL - 2353

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 776

EP - 790

BT - Computer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings

PB - Springer Verlag

ER -