Speed up in Computation of HMAX Features for Object Localization

Takuya Minagawa, Hideo Saito

Research output: Contribution to journalArticle

Abstract

While HMAX features have been proved to have excellent performance in image categorization tasks, the computational cost of recognition is expensive. If we aim to apply the HMAX features to object localization tasks, in which the categorization tasks are repeatedly performed by sliding windows, their processing time increases enormously. In this paper, we propose a method for speed up in computation of object localization based on HMAX features. We found that the HMAX features cause specific redundancies in the sliding window approach. The speed up is achieved by eliminating the redundancies in our method. The results from experiments using the University of Illinois-Urbana-Champaign (UIUC) car dataset and the face detection dataset benchmark (FDDB) indicate that this modification improved processing speeds significantly with insignificant reductions in precision.

Original languageEnglish
Pages (from-to)60-73
Number of pages14
JournalITE Transactions on Media Technology and Applications
Volume2
Issue number1
Publication statusPublished - 2014

Fingerprint

Redundancy
Processing
Face recognition
Railroad cars
Costs
Experiments

Keywords

  • Bags-of-features
  • Feature descriptor
  • HMAX
  • Object localization
  • Object recognition
  • Speed up

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Signal Processing
  • Media Technology

Cite this

Speed up in Computation of HMAX Features for Object Localization. / Minagawa, Takuya; Saito, Hideo.

In: ITE Transactions on Media Technology and Applications, Vol. 2, No. 1, 2014, p. 60-73.

Research output: Contribution to journalArticle

@article{8a2e4570d8e74e7f8362082052c0f8d5,
title = "Speed up in Computation of HMAX Features for Object Localization",
abstract = "While HMAX features have been proved to have excellent performance in image categorization tasks, the computational cost of recognition is expensive. If we aim to apply the HMAX features to object localization tasks, in which the categorization tasks are repeatedly performed by sliding windows, their processing time increases enormously. In this paper, we propose a method for speed up in computation of object localization based on HMAX features. We found that the HMAX features cause specific redundancies in the sliding window approach. The speed up is achieved by eliminating the redundancies in our method. The results from experiments using the University of Illinois-Urbana-Champaign (UIUC) car dataset and the face detection dataset benchmark (FDDB) indicate that this modification improved processing speeds significantly with insignificant reductions in precision.",
keywords = "Bags-of-features, Feature descriptor, HMAX, Object localization, Object recognition, Speed up",
author = "Takuya Minagawa and Hideo Saito",
year = "2014",
language = "English",
volume = "2",
pages = "60--73",
journal = "ITE Transactions on Media Technology and Applications",
issn = "2186-7364",
publisher = "Institute of Image Information and Television Engineers",
number = "1",

}

TY - JOUR

T1 - Speed up in Computation of HMAX Features for Object Localization

AU - Minagawa, Takuya

AU - Saito, Hideo

PY - 2014

Y1 - 2014

N2 - While HMAX features have been proved to have excellent performance in image categorization tasks, the computational cost of recognition is expensive. If we aim to apply the HMAX features to object localization tasks, in which the categorization tasks are repeatedly performed by sliding windows, their processing time increases enormously. In this paper, we propose a method for speed up in computation of object localization based on HMAX features. We found that the HMAX features cause specific redundancies in the sliding window approach. The speed up is achieved by eliminating the redundancies in our method. The results from experiments using the University of Illinois-Urbana-Champaign (UIUC) car dataset and the face detection dataset benchmark (FDDB) indicate that this modification improved processing speeds significantly with insignificant reductions in precision.

AB - While HMAX features have been proved to have excellent performance in image categorization tasks, the computational cost of recognition is expensive. If we aim to apply the HMAX features to object localization tasks, in which the categorization tasks are repeatedly performed by sliding windows, their processing time increases enormously. In this paper, we propose a method for speed up in computation of object localization based on HMAX features. We found that the HMAX features cause specific redundancies in the sliding window approach. The speed up is achieved by eliminating the redundancies in our method. The results from experiments using the University of Illinois-Urbana-Champaign (UIUC) car dataset and the face detection dataset benchmark (FDDB) indicate that this modification improved processing speeds significantly with insignificant reductions in precision.

KW - Bags-of-features

KW - Feature descriptor

KW - HMAX

KW - Object localization

KW - Object recognition

KW - Speed up

UR - http://www.scopus.com/inward/record.url?scp=84951936275&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84951936275&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84951936275

VL - 2

SP - 60

EP - 73

JO - ITE Transactions on Media Technology and Applications

JF - ITE Transactions on Media Technology and Applications

SN - 2186-7364

IS - 1

ER -