Discriminative Histogram Intersection Metric Learning and Its Applications

Peng Yi Hao, Yang Xia, Xiao Xin Li, Seiichiro Kamata, Sheng Yong Chen

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

In this paper, a novel method called discriminative histogram intersection metric learning (DHIML) is proposed for pair matching and classification. Specifically, we introduce a discrimination term for learning a metric from binary information such as same/not-same or similar/dissimilar, and then combine it with the classification error for the discrimination in classifier construction. Compared with conventional approaches, the proposed method has several advantages. 1) The histogram intersection strategy is adopted into metric learning to deal with the widely used histogram features effectively. 2) By introducing discriminative term and classification error term into metric learning, a more discriminative distance metric and a classifier can be learned together. 3) The objective function is robust to outliers and noises for both features and labels in the training. The performance of the proposed method is tested on four applications: face verification, face-track identification, face-track clustering, and image classification. Evaluations on the challenging restricted protocol of Labeled Faces in the Wild (LFW) benchmark, a dataset with more than 7 000 face-tracks, and Caltech-101 dataset validate the robustness and discriminability of the proposed metric learning, compared with the recent state-of-the-art approaches.

Original languageEnglish
Pages (from-to)507-519
Number of pages13
JournalJournal of Computer Science and Technology
Volume32
Issue number3
DOIs
Publication statusPublished - 2017 May 1

Fingerprint

Histogram
Intersection
Metric
Face
Classifiers
Image classification
Discrimination
Classifier
Labels
Distance Metric
Image Classification
Network protocols
Term
Error term
Outlier
Objective function
Learning
Clustering
Binary
Benchmark

Keywords

  • face verification
  • image classification
  • metric learning
  • pair matching

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Software
  • Hardware and Architecture
  • Computer Science Applications
  • Computational Theory and Mathematics

Cite this

Discriminative Histogram Intersection Metric Learning and Its Applications. / Hao, Peng Yi; Xia, Yang; Li, Xiao Xin; Kamata, Seiichiro; Chen, Sheng Yong.

In: Journal of Computer Science and Technology, Vol. 32, No. 3, 01.05.2017, p. 507-519.

Research output: Contribution to journalArticle

Hao, Peng Yi ; Xia, Yang ; Li, Xiao Xin ; Kamata, Seiichiro ; Chen, Sheng Yong. / Discriminative Histogram Intersection Metric Learning and Its Applications. In: Journal of Computer Science and Technology. 2017 ; Vol. 32, No. 3. pp. 507-519.
@article{3278c4869f61437e95906958ba9379bf,
title = "Discriminative Histogram Intersection Metric Learning and Its Applications",
abstract = "In this paper, a novel method called discriminative histogram intersection metric learning (DHIML) is proposed for pair matching and classification. Specifically, we introduce a discrimination term for learning a metric from binary information such as same/not-same or similar/dissimilar, and then combine it with the classification error for the discrimination in classifier construction. Compared with conventional approaches, the proposed method has several advantages. 1) The histogram intersection strategy is adopted into metric learning to deal with the widely used histogram features effectively. 2) By introducing discriminative term and classification error term into metric learning, a more discriminative distance metric and a classifier can be learned together. 3) The objective function is robust to outliers and noises for both features and labels in the training. The performance of the proposed method is tested on four applications: face verification, face-track identification, face-track clustering, and image classification. Evaluations on the challenging restricted protocol of Labeled Faces in the Wild (LFW) benchmark, a dataset with more than 7 000 face-tracks, and Caltech-101 dataset validate the robustness and discriminability of the proposed metric learning, compared with the recent state-of-the-art approaches.",
keywords = "face verification, image classification, metric learning, pair matching",
author = "Hao, {Peng Yi} and Yang Xia and Li, {Xiao Xin} and Seiichiro Kamata and Chen, {Sheng Yong}",
year = "2017",
month = "5",
day = "1",
doi = "10.1007/s11390-017-1740-0",
language = "English",
volume = "32",
pages = "507--519",
journal = "Journal of Computer Science and Technology",
issn = "1000-9000",
publisher = "Springer New York",
number = "3",

}

TY - JOUR

T1 - Discriminative Histogram Intersection Metric Learning and Its Applications

AU - Hao, Peng Yi

AU - Xia, Yang

AU - Li, Xiao Xin

AU - Kamata, Seiichiro

AU - Chen, Sheng Yong

PY - 2017/5/1

Y1 - 2017/5/1

N2 - In this paper, a novel method called discriminative histogram intersection metric learning (DHIML) is proposed for pair matching and classification. Specifically, we introduce a discrimination term for learning a metric from binary information such as same/not-same or similar/dissimilar, and then combine it with the classification error for the discrimination in classifier construction. Compared with conventional approaches, the proposed method has several advantages. 1) The histogram intersection strategy is adopted into metric learning to deal with the widely used histogram features effectively. 2) By introducing discriminative term and classification error term into metric learning, a more discriminative distance metric and a classifier can be learned together. 3) The objective function is robust to outliers and noises for both features and labels in the training. The performance of the proposed method is tested on four applications: face verification, face-track identification, face-track clustering, and image classification. Evaluations on the challenging restricted protocol of Labeled Faces in the Wild (LFW) benchmark, a dataset with more than 7 000 face-tracks, and Caltech-101 dataset validate the robustness and discriminability of the proposed metric learning, compared with the recent state-of-the-art approaches.

AB - In this paper, a novel method called discriminative histogram intersection metric learning (DHIML) is proposed for pair matching and classification. Specifically, we introduce a discrimination term for learning a metric from binary information such as same/not-same or similar/dissimilar, and then combine it with the classification error for the discrimination in classifier construction. Compared with conventional approaches, the proposed method has several advantages. 1) The histogram intersection strategy is adopted into metric learning to deal with the widely used histogram features effectively. 2) By introducing discriminative term and classification error term into metric learning, a more discriminative distance metric and a classifier can be learned together. 3) The objective function is robust to outliers and noises for both features and labels in the training. The performance of the proposed method is tested on four applications: face verification, face-track identification, face-track clustering, and image classification. Evaluations on the challenging restricted protocol of Labeled Faces in the Wild (LFW) benchmark, a dataset with more than 7 000 face-tracks, and Caltech-101 dataset validate the robustness and discriminability of the proposed metric learning, compared with the recent state-of-the-art approaches.

KW - face verification

KW - image classification

KW - metric learning

KW - pair matching

UR - http://www.scopus.com/inward/record.url?scp=85019083977&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85019083977&partnerID=8YFLogxK

U2 - 10.1007/s11390-017-1740-0

DO - 10.1007/s11390-017-1740-0

M3 - Article

VL - 32

SP - 507

EP - 519

JO - Journal of Computer Science and Technology

JF - Journal of Computer Science and Technology

SN - 1000-9000

IS - 3

ER -