Multiple class multiple-instance learning and its application to image categorization

Xinyu Xu, Baoxin Li

Research output: Contribution to journalArticle

Abstract

We propose a Multiple Class Multiple-Instance (MCMI) learning approach and demonstrate its application to the problem of image categorization. Our method extends the binary Multiple-Instance learning approach for image categorization. Instead of constructing a set of binary classifiers (each trained to separate one category from the rest) and then making the final decision based on the winner of all the binary classifiers, our method directly allows the computation of a multi-class classifier by first projecting each training image onto a multi-class feature space and then simultaneously minimizing the multi-class objective function in a Support Vector Machine framework. The multi-class feature space is constructed based on the instance prototypes obtained by Multiple-Instance learning which treats an image as a set of instances with training labels being associated with images rather than instances. The experiment results on two challenging data sets demonstrate that our method achieved better classification accuracy and is less sensitive to the training sample size compared with traditional one-versus-the-rest binary MI classification methods.

Original languageEnglish (US)
Pages (from-to)427-444
Number of pages18
JournalInternational Journal of Image and Graphics
Volume7
Issue number3
DOIs
StatePublished - Jul 1 2007

Fingerprint

Classifiers
Support vector machines
Labels
Experiments

Keywords

  • Image categorization
  • multi-class classification
  • multiple-instance learning
  • support vector machines

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design

Cite this

Multiple class multiple-instance learning and its application to image categorization. / Xu, Xinyu; Li, Baoxin.

In: International Journal of Image and Graphics, Vol. 7, No. 3, 01.07.2007, p. 427-444.

Research output: Contribution to journalArticle

@article{b50c44457a1e4ad4ae0ddeb885fdcafa,
title = "Multiple class multiple-instance learning and its application to image categorization",
abstract = "We propose a Multiple Class Multiple-Instance (MCMI) learning approach and demonstrate its application to the problem of image categorization. Our method extends the binary Multiple-Instance learning approach for image categorization. Instead of constructing a set of binary classifiers (each trained to separate one category from the rest) and then making the final decision based on the winner of all the binary classifiers, our method directly allows the computation of a multi-class classifier by first projecting each training image onto a multi-class feature space and then simultaneously minimizing the multi-class objective function in a Support Vector Machine framework. The multi-class feature space is constructed based on the instance prototypes obtained by Multiple-Instance learning which treats an image as a set of instances with training labels being associated with images rather than instances. The experiment results on two challenging data sets demonstrate that our method achieved better classification accuracy and is less sensitive to the training sample size compared with traditional one-versus-the-rest binary MI classification methods.",
keywords = "Image categorization, multi-class classification, multiple-instance learning, support vector machines",
author = "Xinyu Xu and Baoxin Li",
year = "2007",
month = "7",
day = "1",
doi = "10.1142/S021946780700274X",
language = "English (US)",
volume = "7",
pages = "427--444",
journal = "International Journal of Image and Graphics",
issn = "0219-4678",
publisher = "World Scientific Publishing Co. Pte Ltd",
number = "3",

}

TY - JOUR

T1 - Multiple class multiple-instance learning and its application to image categorization

AU - Xu, Xinyu

AU - Li, Baoxin

PY - 2007/7/1

Y1 - 2007/7/1

N2 - We propose a Multiple Class Multiple-Instance (MCMI) learning approach and demonstrate its application to the problem of image categorization. Our method extends the binary Multiple-Instance learning approach for image categorization. Instead of constructing a set of binary classifiers (each trained to separate one category from the rest) and then making the final decision based on the winner of all the binary classifiers, our method directly allows the computation of a multi-class classifier by first projecting each training image onto a multi-class feature space and then simultaneously minimizing the multi-class objective function in a Support Vector Machine framework. The multi-class feature space is constructed based on the instance prototypes obtained by Multiple-Instance learning which treats an image as a set of instances with training labels being associated with images rather than instances. The experiment results on two challenging data sets demonstrate that our method achieved better classification accuracy and is less sensitive to the training sample size compared with traditional one-versus-the-rest binary MI classification methods.

AB - We propose a Multiple Class Multiple-Instance (MCMI) learning approach and demonstrate its application to the problem of image categorization. Our method extends the binary Multiple-Instance learning approach for image categorization. Instead of constructing a set of binary classifiers (each trained to separate one category from the rest) and then making the final decision based on the winner of all the binary classifiers, our method directly allows the computation of a multi-class classifier by first projecting each training image onto a multi-class feature space and then simultaneously minimizing the multi-class objective function in a Support Vector Machine framework. The multi-class feature space is constructed based on the instance prototypes obtained by Multiple-Instance learning which treats an image as a set of instances with training labels being associated with images rather than instances. The experiment results on two challenging data sets demonstrate that our method achieved better classification accuracy and is less sensitive to the training sample size compared with traditional one-versus-the-rest binary MI classification methods.

KW - Image categorization

KW - multi-class classification

KW - multiple-instance learning

KW - support vector machines

UR - http://www.scopus.com/inward/record.url?scp=44349138350&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=44349138350&partnerID=8YFLogxK

U2 - 10.1142/S021946780700274X

DO - 10.1142/S021946780700274X

M3 - Article

VL - 7

SP - 427

EP - 444

JO - International Journal of Image and Graphics

JF - International Journal of Image and Graphics

SN - 0219-4678

IS - 3

ER -