Improved HOG Descriptors in Image Classification with CP Demonstration

Tan Vo, Dat TRAN, Wanli MA, Khoa NGUYEN

Research output: A Conference proceeding or a Chapter in BookConference contribution

2 Citations (Scopus)

Abstract

Histogram of Oriented Gradients (HOG) has been widely used in computer vision as feature descriptors for detecting objects in scenes. We present in this paper a new approach to HOG in image classification that will provide an opportunity to explore new ways to improve the effectiveness of HOG image descriptors. We investigate applying tensor decomposition on HOG descriptors then using them as image features to build image models using support vector machine. The aim of this approach is to produce a more robust and compact version of HOG features. An image classification experiment is performed to evaluate the effectiveness of this approach as well as to identify all ideal parameter values involved. Experimental results show a good improvement in image classification rate for the proposed approach.
Original languageEnglish
Title of host publicationInternational Conference on Neural Information Processing (ICONIP 2013)
Subtitle of host publicationLecture Notes in Computer Science
EditorsMinho Lee, Akira Hirose, Zeng-Guang Hou, Rhee Man Kil
Place of PublicationGermany
PublisherSpringer
Pages384-391
Number of pages8
Volume8228
ISBN (Electronic)9783642420511
ISBN (Print)9783642420504
DOIs
Publication statusPublished - 2013
Event20th International Conference on Neural Information Processing (ICONIP 2013) - Daegu, Daegu, Korea, Republic of
Duration: 3 Nov 20137 Nov 2013

Conference

Conference20th International Conference on Neural Information Processing (ICONIP 2013)
Abbreviated titleICONIP 2013
CountryKorea, Republic of
CityDaegu
Period3/11/137/11/13

Fingerprint

Image classification
Demonstrations
Computer vision
Tensors
Support vector machines
Decomposition
Experiments

Cite this

Vo, T., TRAN, D., MA, W., & NGUYEN, K. (2013). Improved HOG Descriptors in Image Classification with CP Demonstration. In M. Lee, A. Hirose, Z-G. Hou, & R. M. Kil (Eds.), International Conference on Neural Information Processing (ICONIP 2013): Lecture Notes in Computer Science (Vol. 8228, pp. 384-391). Germany: Springer. https://doi.org/10.1007/978-3-642-42051-1_48
Vo, Tan ; TRAN, Dat ; MA, Wanli ; NGUYEN, Khoa. / Improved HOG Descriptors in Image Classification with CP Demonstration. International Conference on Neural Information Processing (ICONIP 2013): Lecture Notes in Computer Science. editor / Minho Lee ; Akira Hirose ; Zeng-Guang Hou ; Rhee Man Kil. Vol. 8228 Germany : Springer, 2013. pp. 384-391
@inproceedings{4cc43f3cb40346a0a3e2745cced591cb,
title = "Improved HOG Descriptors in Image Classification with CP Demonstration",
abstract = "Histogram of Oriented Gradients (HOG) has been widely used in computer vision as feature descriptors for detecting objects in scenes. We present in this paper a new approach to HOG in image classification that will provide an opportunity to explore new ways to improve the effectiveness of HOG image descriptors. We investigate applying tensor decomposition on HOG descriptors then using them as image features to build image models using support vector machine. The aim of this approach is to produce a more robust and compact version of HOG features. An image classification experiment is performed to evaluate the effectiveness of this approach as well as to identify all ideal parameter values involved. Experimental results show a good improvement in image classification rate for the proposed approach.",
keywords = "Age and Gender Classification, Tensor Decomposition, histogram of oriented gradients",
author = "Tan Vo and Dat TRAN and Wanli MA and Khoa NGUYEN",
year = "2013",
doi = "10.1007/978-3-642-42051-1_48",
language = "English",
isbn = "9783642420504",
volume = "8228",
pages = "384--391",
editor = "Minho Lee and Akira Hirose and Zeng-Guang Hou and Kil, {Rhee Man}",
booktitle = "International Conference on Neural Information Processing (ICONIP 2013)",
publisher = "Springer",
address = "Netherlands",

}

Vo, T, TRAN, D, MA, W & NGUYEN, K 2013, Improved HOG Descriptors in Image Classification with CP Demonstration. in M Lee, A Hirose, Z-G Hou & RM Kil (eds), International Conference on Neural Information Processing (ICONIP 2013): Lecture Notes in Computer Science. vol. 8228, Springer, Germany, pp. 384-391, 20th International Conference on Neural Information Processing (ICONIP 2013), Daegu, Korea, Republic of, 3/11/13. https://doi.org/10.1007/978-3-642-42051-1_48

Improved HOG Descriptors in Image Classification with CP Demonstration. / Vo, Tan; TRAN, Dat; MA, Wanli; NGUYEN, Khoa.

International Conference on Neural Information Processing (ICONIP 2013): Lecture Notes in Computer Science. ed. / Minho Lee; Akira Hirose; Zeng-Guang Hou; Rhee Man Kil. Vol. 8228 Germany : Springer, 2013. p. 384-391.

Research output: A Conference proceeding or a Chapter in BookConference contribution

TY - GEN

T1 - Improved HOG Descriptors in Image Classification with CP Demonstration

AU - Vo, Tan

AU - TRAN, Dat

AU - MA, Wanli

AU - NGUYEN, Khoa

PY - 2013

Y1 - 2013

N2 - Histogram of Oriented Gradients (HOG) has been widely used in computer vision as feature descriptors for detecting objects in scenes. We present in this paper a new approach to HOG in image classification that will provide an opportunity to explore new ways to improve the effectiveness of HOG image descriptors. We investigate applying tensor decomposition on HOG descriptors then using them as image features to build image models using support vector machine. The aim of this approach is to produce a more robust and compact version of HOG features. An image classification experiment is performed to evaluate the effectiveness of this approach as well as to identify all ideal parameter values involved. Experimental results show a good improvement in image classification rate for the proposed approach.

AB - Histogram of Oriented Gradients (HOG) has been widely used in computer vision as feature descriptors for detecting objects in scenes. We present in this paper a new approach to HOG in image classification that will provide an opportunity to explore new ways to improve the effectiveness of HOG image descriptors. We investigate applying tensor decomposition on HOG descriptors then using them as image features to build image models using support vector machine. The aim of this approach is to produce a more robust and compact version of HOG features. An image classification experiment is performed to evaluate the effectiveness of this approach as well as to identify all ideal parameter values involved. Experimental results show a good improvement in image classification rate for the proposed approach.

KW - Age and Gender Classification

KW - Tensor Decomposition

KW - histogram of oriented gradients

UR - https://link.springer.com/chapter/10.1007/978-3-642-42051-1_48

U2 - 10.1007/978-3-642-42051-1_48

DO - 10.1007/978-3-642-42051-1_48

M3 - Conference contribution

SN - 9783642420504

VL - 8228

SP - 384

EP - 391

BT - International Conference on Neural Information Processing (ICONIP 2013)

A2 - Lee, Minho

A2 - Hirose, Akira

A2 - Hou, Zeng-Guang

A2 - Kil, Rhee Man

PB - Springer

CY - Germany

ER -

Vo T, TRAN D, MA W, NGUYEN K. Improved HOG Descriptors in Image Classification with CP Demonstration. In Lee M, Hirose A, Hou Z-G, Kil RM, editors, International Conference on Neural Information Processing (ICONIP 2013): Lecture Notes in Computer Science. Vol. 8228. Germany: Springer. 2013. p. 384-391 https://doi.org/10.1007/978-3-642-42051-1_48