Learning invariant structure for object identification by using graph methods

Bai Xiao, Yi Zhe Song, Peter Hall

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

The problem of learning the class identity of visual objects has received considerable attention recently. With rare exception, all of the work to date assumes low variation in appearance, which limits them to a single depictive style usually photographic. The same object depicted in other styles - as a drawing, perhaps - cannot be identified reliably. Yet humans are able to name the object no matter how it is depicted, and even recognize a real object having previously seen only a drawing. This paper describes a classifier which is unique in being able to learn class identity no matter how the class instances are depicted. The key to this is our proposition that topological structure is a class invariant. Practically, we depend on spectral graph analysis of a hierarchical description of an image to construct a feature vector of fixed dimension. Hence structure is transformed to a feature vector, which can be classified using standard methods. We demonstrate the classifier on several diverse classes.
Original languageEnglish
Pages (from-to)1023-1031
Number of pages9
JournalComputer Vision and Image Understanding
Volume115
Issue number7
DOIs
Publication statusPublished - 2011

Fingerprint

Classifiers

Cite this

Learning invariant structure for object identification by using graph methods. / Xiao, Bai; Song, Yi Zhe; Hall, Peter.

In: Computer Vision and Image Understanding, Vol. 115, No. 7, 2011, p. 1023-1031.

Research output: Contribution to journalArticle

@article{4cca42931dc14a8c87c225990c83077c,
title = "Learning invariant structure for object identification by using graph methods",
abstract = "The problem of learning the class identity of visual objects has received considerable attention recently. With rare exception, all of the work to date assumes low variation in appearance, which limits them to a single depictive style usually photographic. The same object depicted in other styles - as a drawing, perhaps - cannot be identified reliably. Yet humans are able to name the object no matter how it is depicted, and even recognize a real object having previously seen only a drawing. This paper describes a classifier which is unique in being able to learn class identity no matter how the class instances are depicted. The key to this is our proposition that topological structure is a class invariant. Practically, we depend on spectral graph analysis of a hierarchical description of an image to construct a feature vector of fixed dimension. Hence structure is transformed to a feature vector, which can be classified using standard methods. We demonstrate the classifier on several diverse classes.",
author = "Bai Xiao and Song, {Yi Zhe} and Peter Hall",
year = "2011",
doi = "10.1016/j.cviu.2010.12.016",
language = "English",
volume = "115",
pages = "1023--1031",
journal = "Computer Vision and Image Understanding",
issn = "1077-3142",
publisher = "Elsevier Academic Press Inc",
number = "7",

}

TY - JOUR

T1 - Learning invariant structure for object identification by using graph methods

AU - Xiao, Bai

AU - Song, Yi Zhe

AU - Hall, Peter

PY - 2011

Y1 - 2011

N2 - The problem of learning the class identity of visual objects has received considerable attention recently. With rare exception, all of the work to date assumes low variation in appearance, which limits them to a single depictive style usually photographic. The same object depicted in other styles - as a drawing, perhaps - cannot be identified reliably. Yet humans are able to name the object no matter how it is depicted, and even recognize a real object having previously seen only a drawing. This paper describes a classifier which is unique in being able to learn class identity no matter how the class instances are depicted. The key to this is our proposition that topological structure is a class invariant. Practically, we depend on spectral graph analysis of a hierarchical description of an image to construct a feature vector of fixed dimension. Hence structure is transformed to a feature vector, which can be classified using standard methods. We demonstrate the classifier on several diverse classes.

AB - The problem of learning the class identity of visual objects has received considerable attention recently. With rare exception, all of the work to date assumes low variation in appearance, which limits them to a single depictive style usually photographic. The same object depicted in other styles - as a drawing, perhaps - cannot be identified reliably. Yet humans are able to name the object no matter how it is depicted, and even recognize a real object having previously seen only a drawing. This paper describes a classifier which is unique in being able to learn class identity no matter how the class instances are depicted. The key to this is our proposition that topological structure is a class invariant. Practically, we depend on spectral graph analysis of a hierarchical description of an image to construct a feature vector of fixed dimension. Hence structure is transformed to a feature vector, which can be classified using standard methods. We demonstrate the classifier on several diverse classes.

UR - http://www.scopus.com/inward/record.url?scp=79956157435&partnerID=8YFLogxK

UR - http://dx.doi.org/10.1016/j.cviu.2010.12.016

U2 - 10.1016/j.cviu.2010.12.016

DO - 10.1016/j.cviu.2010.12.016

M3 - Article

VL - 115

SP - 1023

EP - 1031

JO - Computer Vision and Image Understanding

JF - Computer Vision and Image Understanding

SN - 1077-3142

IS - 7

ER -