Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification

Alper Ulu; Hazim Kemal Ekenel

Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification

Alper Ulu^*, Hazim Kemal Ekenel

^*Corresponding author for this work

Department of Computer Engineering

Istanbul Technical University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Video surveillance systems have great importance to ensure public safety. Today, these kind of systems not only capture and distribute video but also have various smart applications. Person re-identification is one of the most important of these applications. In this work, we have exploited deep convolutional neural networkbased representations for cross dataset person re-identification problem. We have selected well-known deep convolutional neural network models, namely, AlexNet, VGG-16, and GoogLeNet, and fine-tuned them with the largest publicly available person re-identification datasets. We have employed cosine similarity metric to calculate the similarity between extracted features. CUHK03 and Market-1501 datasets were used as the training sets and the proposed method has been tested on the VIPeR dataset. Superior results have been obtained with the proposed method compared to the state-of-the-art methods in the field.

Original language	English
Title of host publication	VISAPP
Editors	Alain Tremeau, Jose Braz, Francisco Imai
Publisher	SciTePress
Pages	571-578
Number of pages	8
ISBN (Electronic)	9789897582257
Publication status	Published - 2017
Event	12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2017 - Porto, Portugal Duration: 27 Feb 2017 → 1 Mar 2017

Publication series

Name	VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications
Volume	4

Conference

Conference	12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2017
Country/Territory	Portugal
City	Porto
Period	27/02/17 → 1/03/17

Bibliographical note

Publisher Copyright:
© 2017 by SCITEPRESS - Science and Technology Publications, Lda.

Funding

This work was supported by TUBITAK project number 113E067, by a Marie Curie FP7 Integration Grant within the 7th EU Framework Programme, and by Istanbul Technical University Research Fund project number 39634.

Funders	Funder number
7th EU Framework Programme
Istanbul Technical University Research Fund	39634
TUBITAK	113E067
Marie Curie

Keywords

Convolutional Neural Networks
Deep Learning
Person Re-identification

Cite this

Ulu, A., & Ekenel, H. K. (2017). Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification. In A. Tremeau, J. Braz, & F. Imai (Eds.), VISAPP (pp. 571-578). (VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications; Vol. 4). SciTePress.

Ulu, Alper ; Ekenel, Hazim Kemal. / Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification. VISAPP. editor / Alain Tremeau ; Jose Braz ; Francisco Imai. SciTePress, 2017. pp. 571-578 (VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications).

@inproceedings{15e617aa409b405aba88a9ed2b0f0975,

title = "Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification",

abstract = "Video surveillance systems have great importance to ensure public safety. Today, these kind of systems not only capture and distribute video but also have various smart applications. Person re-identification is one of the most important of these applications. In this work, we have exploited deep convolutional neural networkbased representations for cross dataset person re-identification problem. We have selected well-known deep convolutional neural network models, namely, AlexNet, VGG-16, and GoogLeNet, and fine-tuned them with the largest publicly available person re-identification datasets. We have employed cosine similarity metric to calculate the similarity between extracted features. CUHK03 and Market-1501 datasets were used as the training sets and the proposed method has been tested on the VIPeR dataset. Superior results have been obtained with the proposed method compared to the state-of-the-art methods in the field.",

keywords = "Convolutional Neural Networks, Deep Learning, Person Re-identification",

author = "Alper Ulu and Ekenel, {Hazim Kemal}",

note = "Publisher Copyright: {\textcopyright} 2017 by SCITEPRESS - Science and Technology Publications, Lda.; 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2017 ; Conference date: 27-02-2017 Through 01-03-2017",

year = "2017",

language = "English",

series = "VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications",

publisher = "SciTePress",

pages = "571--578",

editor = "Alain Tremeau and Jose Braz and Francisco Imai",

booktitle = "VISAPP",

}

Ulu, A & Ekenel, HK 2017, Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification. in A Tremeau, J Braz & F Imai (eds), VISAPP. VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, vol. 4, SciTePress, pp. 571-578, 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2017, Porto, Portugal, 27/02/17.

Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification. / Ulu, Alper; Ekenel, Hazim Kemal.
VISAPP. ed. / Alain Tremeau; Jose Braz; Francisco Imai. SciTePress, 2017. p. 571-578 (VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications; Vol. 4).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification

AU - Ulu, Alper

AU - Ekenel, Hazim Kemal

PY - 2017

Y1 - 2017

N2 - Video surveillance systems have great importance to ensure public safety. Today, these kind of systems not only capture and distribute video but also have various smart applications. Person re-identification is one of the most important of these applications. In this work, we have exploited deep convolutional neural networkbased representations for cross dataset person re-identification problem. We have selected well-known deep convolutional neural network models, namely, AlexNet, VGG-16, and GoogLeNet, and fine-tuned them with the largest publicly available person re-identification datasets. We have employed cosine similarity metric to calculate the similarity between extracted features. CUHK03 and Market-1501 datasets were used as the training sets and the proposed method has been tested on the VIPeR dataset. Superior results have been obtained with the proposed method compared to the state-of-the-art methods in the field.

AB - Video surveillance systems have great importance to ensure public safety. Today, these kind of systems not only capture and distribute video but also have various smart applications. Person re-identification is one of the most important of these applications. In this work, we have exploited deep convolutional neural networkbased representations for cross dataset person re-identification problem. We have selected well-known deep convolutional neural network models, namely, AlexNet, VGG-16, and GoogLeNet, and fine-tuned them with the largest publicly available person re-identification datasets. We have employed cosine similarity metric to calculate the similarity between extracted features. CUHK03 and Market-1501 datasets were used as the training sets and the proposed method has been tested on the VIPeR dataset. Superior results have been obtained with the proposed method compared to the state-of-the-art methods in the field.

KW - Convolutional Neural Networks

KW - Deep Learning

KW - Person Re-identification

UR - http://www.scopus.com/inward/record.url?scp=85047874229&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85047874229

T3 - VISIGRAPP 2017 - Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications

SP - 571

EP - 578

BT - VISAPP

A2 - Tremeau, Alain

A2 - Braz, Jose

A2 - Imai, Francisco

PB - SciTePress

T2 - 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2017

Y2 - 27 February 2017 through 1 March 2017

ER -

Evaluation of deep convolutional neural network-based representations for cross dataset person re-identification

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Other files and links

Fingerprint

Cite this