Adjustment learning and relevant component analysis

Noam Shental, Tomer Hertz, Daphna Weinshall, Misha Pavel

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

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
Title of host publicationComputer Vision - ECCV 2002 - 7th European Conference on Computer Vision, Proceedings
EditorsAnders Heyden, Gunnar Sparr, Mads Nielsen, Peter Johansen
PublisherSpringer Verlag
Pages776-790
Number of pages15
Volume2353
ISBN (Electronic)9783540437482
DOIs
StatePublished - 2002
Externally publishedYes
Event7th European Conference on Computer Vision, ECCV 2002 - Copenhagen, Denmark
Duration: 28 May 200231 May 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2353
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference7th European Conference on Computer Vision, ECCV 2002
Country/TerritoryDenmark
CityCopenhagen
Period28/05/0231/05/02

Bibliographical note

Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 2002.

Fingerprint

Dive into the research topics of 'Adjustment learning and relevant component analysis'. Together they form a unique fingerprint.

Cite this