Region-based automatic web image selection

Keiji Yanai, Jacobus J Barnard

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

2 Citations (Scopus)

Abstract

We propose a new Web image selection method which employs the region-based bag-of-features representation. The contribution of this work is (1) to introduce the region-based bag-of-features representation into an Web image selection task where training data is incomplete, and (2) to prove its effectiveness by experiments with both generative and discriminative machine learning methods. In the experiments, we used a multiple-instance learning SVM and a standard SVM as discriminative methods, and pLSA and LDA mixture models as probabilistic generative methods. Several works on Web image filtering task with bag-of-features have been proposed so far. However, in case that the training data includes much noise, sufficient results could not be obtained. In this paper, we divide images into regions and classify each region instead of classifying whole images. By this region-based classification, we can separate foreground regions from background regions and achieve more effective image training from incomplete training data. By the experiments, we show that the results by the proposed methods outperformed the results by the whole-image-based bag-of-features.

Original languageEnglish (US)
Title of host publicationMIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval
Pages305-312
Number of pages8
DOIs
StatePublished - 2010
Event2010 ACM SIGMM International Conference on Multimedia Information Retrieval, MIR 2010 - Philadelphia, PA, United States
Duration: Mar 29 2010Mar 31 2010

Other

Other2010 ACM SIGMM International Conference on Multimedia Information Retrieval, MIR 2010
CountryUnited States
CityPhiladelphia, PA
Period3/29/103/31/10

Fingerprint

Experiments
Learning systems

Keywords

  • LDA
  • Multiple instance learning
  • PLSA
  • Region-based
  • SVM
  • Web image mining

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Information Systems

Cite this

Yanai, K., & Barnard, J. J. (2010). Region-based automatic web image selection. In MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval (pp. 305-312) https://doi.org/10.1145/1743384.1743436

Region-based automatic web image selection. / Yanai, Keiji; Barnard, Jacobus J.

MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval. 2010. p. 305-312.

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

Yanai, K & Barnard, JJ 2010, Region-based automatic web image selection. in MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval. pp. 305-312, 2010 ACM SIGMM International Conference on Multimedia Information Retrieval, MIR 2010, Philadelphia, PA, United States, 3/29/10. https://doi.org/10.1145/1743384.1743436
Yanai K, Barnard JJ. Region-based automatic web image selection. In MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval. 2010. p. 305-312 https://doi.org/10.1145/1743384.1743436
Yanai, Keiji ; Barnard, Jacobus J. / Region-based automatic web image selection. MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval. 2010. pp. 305-312
@inproceedings{5ff180c407a24cc88dbb823ba41ea440,
title = "Region-based automatic web image selection",
abstract = "We propose a new Web image selection method which employs the region-based bag-of-features representation. The contribution of this work is (1) to introduce the region-based bag-of-features representation into an Web image selection task where training data is incomplete, and (2) to prove its effectiveness by experiments with both generative and discriminative machine learning methods. In the experiments, we used a multiple-instance learning SVM and a standard SVM as discriminative methods, and pLSA and LDA mixture models as probabilistic generative methods. Several works on Web image filtering task with bag-of-features have been proposed so far. However, in case that the training data includes much noise, sufficient results could not be obtained. In this paper, we divide images into regions and classify each region instead of classifying whole images. By this region-based classification, we can separate foreground regions from background regions and achieve more effective image training from incomplete training data. By the experiments, we show that the results by the proposed methods outperformed the results by the whole-image-based bag-of-features.",
keywords = "LDA, Multiple instance learning, PLSA, Region-based, SVM, Web image mining",
author = "Keiji Yanai and Barnard, {Jacobus J}",
year = "2010",
doi = "10.1145/1743384.1743436",
language = "English (US)",
isbn = "9781605588155",
pages = "305--312",
booktitle = "MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval",

}

TY - GEN

T1 - Region-based automatic web image selection

AU - Yanai, Keiji

AU - Barnard, Jacobus J

PY - 2010

Y1 - 2010

N2 - We propose a new Web image selection method which employs the region-based bag-of-features representation. The contribution of this work is (1) to introduce the region-based bag-of-features representation into an Web image selection task where training data is incomplete, and (2) to prove its effectiveness by experiments with both generative and discriminative machine learning methods. In the experiments, we used a multiple-instance learning SVM and a standard SVM as discriminative methods, and pLSA and LDA mixture models as probabilistic generative methods. Several works on Web image filtering task with bag-of-features have been proposed so far. However, in case that the training data includes much noise, sufficient results could not be obtained. In this paper, we divide images into regions and classify each region instead of classifying whole images. By this region-based classification, we can separate foreground regions from background regions and achieve more effective image training from incomplete training data. By the experiments, we show that the results by the proposed methods outperformed the results by the whole-image-based bag-of-features.

AB - We propose a new Web image selection method which employs the region-based bag-of-features representation. The contribution of this work is (1) to introduce the region-based bag-of-features representation into an Web image selection task where training data is incomplete, and (2) to prove its effectiveness by experiments with both generative and discriminative machine learning methods. In the experiments, we used a multiple-instance learning SVM and a standard SVM as discriminative methods, and pLSA and LDA mixture models as probabilistic generative methods. Several works on Web image filtering task with bag-of-features have been proposed so far. However, in case that the training data includes much noise, sufficient results could not be obtained. In this paper, we divide images into regions and classify each region instead of classifying whole images. By this region-based classification, we can separate foreground regions from background regions and achieve more effective image training from incomplete training data. By the experiments, we show that the results by the proposed methods outperformed the results by the whole-image-based bag-of-features.

KW - LDA

KW - Multiple instance learning

KW - PLSA

KW - Region-based

KW - SVM

KW - Web image mining

UR - http://www.scopus.com/inward/record.url?scp=77952363681&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77952363681&partnerID=8YFLogxK

U2 - 10.1145/1743384.1743436

DO - 10.1145/1743384.1743436

M3 - Conference contribution

SN - 9781605588155

SP - 305

EP - 312

BT - MIR 2010 - Proceedings of the 2010 ACM SIGMM International Conference on Multimedia Information Retrieval

ER -