Learning image similarity from flickr groups using fast kernel machines

Research output: Contribution to journalArticle

Abstract

Measuring image similarity is a central topic in computer vision. In this paper, we propose to measure image similarity by learning from the online Flickr image groups. We do so by: Choosing 103 Flickr groups, building a one-versus-all multiclass classifier to classify test images into a group, taking the set of responses of the classifiers as features, calculating the distance between feature vectors to measure image similarity. Experimental results on the Corel dataset and the PASCAL VOC 2007 dataset show that our approach performs better on image matching, retrieval, and classification than using conventional visual features. To build our similarity measure, we need one-versus-all classifiers that are accurate and can be trained quickly on very large quantities of data. We adopt an SVM classifier with a histogram intersection kernel. We describe a novel fast training algorithm for this classifier: the Stochastic Intersection Kernel MAchine (SIKMA) training algorithm. This method can produce a kernel classifier that is more accurate than a linear classifier on tens of thousands of examples in minutes.

Original languageEnglish (US)
Article number6133292
Pages (from-to)2177-2188
Number of pages12
JournalIEEE transactions on pattern analysis and machine intelligence
Volume34
Issue number11
DOIs
StatePublished - Oct 1 2012

Fingerprint

Kernel Machines
Classifiers
Classifier
Training Algorithm
Intersection
kernel
Image matching
Image Matching
Multi-class
Similarity
Learning
Feature Vector
Volatile organic compounds
Similarity Measure
Computer Vision
Histogram
Computer vision
Fast Algorithm
Retrieval
Classify

Keywords

  • Image similarity
  • image classification
  • image organization
  • kernel machines
  • online learning
  • stochastic gradient descent

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition
  • Computational Theory and Mathematics
  • Artificial Intelligence
  • Applied Mathematics

Cite this

Learning image similarity from flickr groups using fast kernel machines. / Wang, Gang; Hoiem, Derek W; Forsyth, David Alexander.

In: IEEE transactions on pattern analysis and machine intelligence, Vol. 34, No. 11, 6133292, 01.10.2012, p. 2177-2188.

Research output: Contribution to journalArticle

@article{740617af0b4741e497af827aa2f7ade7,
title = "Learning image similarity from flickr groups using fast kernel machines",
abstract = "Measuring image similarity is a central topic in computer vision. In this paper, we propose to measure image similarity by learning from the online Flickr image groups. We do so by: Choosing 103 Flickr groups, building a one-versus-all multiclass classifier to classify test images into a group, taking the set of responses of the classifiers as features, calculating the distance between feature vectors to measure image similarity. Experimental results on the Corel dataset and the PASCAL VOC 2007 dataset show that our approach performs better on image matching, retrieval, and classification than using conventional visual features. To build our similarity measure, we need one-versus-all classifiers that are accurate and can be trained quickly on very large quantities of data. We adopt an SVM classifier with a histogram intersection kernel. We describe a novel fast training algorithm for this classifier: the Stochastic Intersection Kernel MAchine (SIKMA) training algorithm. This method can produce a kernel classifier that is more accurate than a linear classifier on tens of thousands of examples in minutes.",
keywords = "Image similarity, image classification, image organization, kernel machines, online learning, stochastic gradient descent",
author = "Gang Wang and Hoiem, {Derek W} and Forsyth, {David Alexander}",
year = "2012",
month = "10",
day = "1",
doi = "10.1109/TPAMI.2012.29",
language = "English (US)",
volume = "34",
pages = "2177--2188",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "11",

}

TY - JOUR

T1 - Learning image similarity from flickr groups using fast kernel machines

AU - Wang, Gang

AU - Hoiem, Derek W

AU - Forsyth, David Alexander

PY - 2012/10/1

Y1 - 2012/10/1

N2 - Measuring image similarity is a central topic in computer vision. In this paper, we propose to measure image similarity by learning from the online Flickr image groups. We do so by: Choosing 103 Flickr groups, building a one-versus-all multiclass classifier to classify test images into a group, taking the set of responses of the classifiers as features, calculating the distance between feature vectors to measure image similarity. Experimental results on the Corel dataset and the PASCAL VOC 2007 dataset show that our approach performs better on image matching, retrieval, and classification than using conventional visual features. To build our similarity measure, we need one-versus-all classifiers that are accurate and can be trained quickly on very large quantities of data. We adopt an SVM classifier with a histogram intersection kernel. We describe a novel fast training algorithm for this classifier: the Stochastic Intersection Kernel MAchine (SIKMA) training algorithm. This method can produce a kernel classifier that is more accurate than a linear classifier on tens of thousands of examples in minutes.

AB - Measuring image similarity is a central topic in computer vision. In this paper, we propose to measure image similarity by learning from the online Flickr image groups. We do so by: Choosing 103 Flickr groups, building a one-versus-all multiclass classifier to classify test images into a group, taking the set of responses of the classifiers as features, calculating the distance between feature vectors to measure image similarity. Experimental results on the Corel dataset and the PASCAL VOC 2007 dataset show that our approach performs better on image matching, retrieval, and classification than using conventional visual features. To build our similarity measure, we need one-versus-all classifiers that are accurate and can be trained quickly on very large quantities of data. We adopt an SVM classifier with a histogram intersection kernel. We describe a novel fast training algorithm for this classifier: the Stochastic Intersection Kernel MAchine (SIKMA) training algorithm. This method can produce a kernel classifier that is more accurate than a linear classifier on tens of thousands of examples in minutes.

KW - Image similarity

KW - image classification

KW - image organization

KW - kernel machines

KW - online learning

KW - stochastic gradient descent

UR - http://www.scopus.com/inward/record.url?scp=84866644789&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84866644789&partnerID=8YFLogxK

U2 - 10.1109/TPAMI.2012.29

DO - 10.1109/TPAMI.2012.29

M3 - Article

C2 - 22997127

AN - SCOPUS:84866644789

VL - 34

SP - 2177

EP - 2188

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 11

M1 - 6133292

ER -