Autoencoder based dimensionality reduction of feature vectors for object recognition

Reyhan Kevser Keser; Behcet Ugur Toreyin

doi:10.1109/SITIS.2019.00097

Autoencoder based dimensionality reduction of feature vectors for object recognition

Reyhan Kevser Keser, Behcet Ugur Toreyin

Informatics Institute

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

3 Citations (Scopus)

Abstract

Object recognition can be performed with high accuracy thanks to the robust feature descriptors defining the significant areas in images. However, these features suffer from high dimensional structure, in other words 'curse of dimensionality' for further processes. Autoencoders (AE) are proposed in this study to solve the dimensionality reduction problem of visual features. To assess the efficacy, object recognition is performed using reduced dimensional visual features. For this purpose, dimensionalities of three well-known feature vectors, namely, HOG, SIFT and SURF, are reduced to half. Moreover, deep learning based features are also reduced. Then, reduced vectors, which are called as AE-HOG, AE-SIFT, AE-SURF and AE-DEEP are fed to object recognition task. Also, dimensionality reduction is implemented by a variant of AE, variational autoencoder (VAE) and PCA, which is the most studied unsupervised method for these features, and the results are compared. Furthermore, all experiments are repeated on noisy images. Results suggest that dimensionality reduction of these feature vectors can be accomplished successfully owing to the proposed method.

Original language	English
Title of host publication	Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019
Editors	Kokou Yetongnon, Albert Dipanda, Gabriella Sanniti di Baja, Luigi Gallo, Richard Chbeir
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	577-584
Number of pages	8
ISBN (Electronic)	9781728156866
DOIs	https://doi.org/10.1109/SITIS.2019.00097
Publication status	Published - Nov 2019
Event	15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019 - Sorrento, Italy Duration: 26 Nov 2019 → 29 Nov 2019

Publication series

Name	Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019

Conference

Conference	15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019
Country/Territory	Italy
City	Sorrento
Period	26/11/19 → 29/11/19

Bibliographical note

Publisher Copyright:
© 2019 IEEE.

Funding

This work is supported in part by Istanbul Technical University (ITU) VodafoneFuture Lab under Project ITUVF20180901P04 and by ITU BAP MGA-2017-40964.

Funders	Funder number
Istanbul Teknik Üniversitesi	BAP MGA-2017-40964, ITUVF20180901P04

Keywords

Autoencoder
Dimensionality reduction
HOG
SIFT
SURF

Access to Document

10.1109/SITIS.2019.00097

Cite this

Keser, R. K., & Toreyin, B. U. (2019). Autoencoder based dimensionality reduction of feature vectors for object recognition. In K. Yetongnon, A. Dipanda, G. Sanniti di Baja, L. Gallo, & R. Chbeir (Eds.), Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019 (pp. 577-584). Article 9067860 (Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SITIS.2019.00097

Keser, Reyhan Kevser ; Toreyin, Behcet Ugur. / Autoencoder based dimensionality reduction of feature vectors for object recognition. Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019. editor / Kokou Yetongnon ; Albert Dipanda ; Gabriella Sanniti di Baja ; Luigi Gallo ; Richard Chbeir. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 577-584 (Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019).

@inproceedings{052982eb1d1e4cff9f0a3c4d8b1e9391,

title = "Autoencoder based dimensionality reduction of feature vectors for object recognition",

abstract = "Object recognition can be performed with high accuracy thanks to the robust feature descriptors defining the significant areas in images. However, these features suffer from high dimensional structure, in other words 'curse of dimensionality' for further processes. Autoencoders (AE) are proposed in this study to solve the dimensionality reduction problem of visual features. To assess the efficacy, object recognition is performed using reduced dimensional visual features. For this purpose, dimensionalities of three well-known feature vectors, namely, HOG, SIFT and SURF, are reduced to half. Moreover, deep learning based features are also reduced. Then, reduced vectors, which are called as AE-HOG, AE-SIFT, AE-SURF and AE-DEEP are fed to object recognition task. Also, dimensionality reduction is implemented by a variant of AE, variational autoencoder (VAE) and PCA, which is the most studied unsupervised method for these features, and the results are compared. Furthermore, all experiments are repeated on noisy images. Results suggest that dimensionality reduction of these feature vectors can be accomplished successfully owing to the proposed method.",

keywords = "Autoencoder, Dimensionality reduction, HOG, SIFT, SURF",

author = "Keser, {Reyhan Kevser} and Toreyin, {Behcet Ugur}",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019 ; Conference date: 26-11-2019 Through 29-11-2019",

year = "2019",

month = nov,

doi = "10.1109/SITIS.2019.00097",

language = "English",

series = "Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "577--584",

editor = "Kokou Yetongnon and Albert Dipanda and {Sanniti di Baja}, Gabriella and Luigi Gallo and Richard Chbeir",

booktitle = "Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019",

address = "United States",

}

Keser, RK & Toreyin, BU 2019, Autoencoder based dimensionality reduction of feature vectors for object recognition. in K Yetongnon, A Dipanda, G Sanniti di Baja, L Gallo & R Chbeir (eds), Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019., 9067860, Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019, Institute of Electrical and Electronics Engineers Inc., pp. 577-584, 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019, Sorrento, Italy, 26/11/19. https://doi.org/10.1109/SITIS.2019.00097

Autoencoder based dimensionality reduction of feature vectors for object recognition. / Keser, Reyhan Kevser ; Toreyin, Behcet Ugur.
Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019. ed. / Kokou Yetongnon; Albert Dipanda; Gabriella Sanniti di Baja; Luigi Gallo; Richard Chbeir. Institute of Electrical and Electronics Engineers Inc., 2019. p. 577-584 9067860 (Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019).

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

TY - GEN

T1 - Autoencoder based dimensionality reduction of feature vectors for object recognition

AU - Keser, Reyhan Kevser

AU - Toreyin, Behcet Ugur

PY - 2019/11

Y1 - 2019/11

N2 - Object recognition can be performed with high accuracy thanks to the robust feature descriptors defining the significant areas in images. However, these features suffer from high dimensional structure, in other words 'curse of dimensionality' for further processes. Autoencoders (AE) are proposed in this study to solve the dimensionality reduction problem of visual features. To assess the efficacy, object recognition is performed using reduced dimensional visual features. For this purpose, dimensionalities of three well-known feature vectors, namely, HOG, SIFT and SURF, are reduced to half. Moreover, deep learning based features are also reduced. Then, reduced vectors, which are called as AE-HOG, AE-SIFT, AE-SURF and AE-DEEP are fed to object recognition task. Also, dimensionality reduction is implemented by a variant of AE, variational autoencoder (VAE) and PCA, which is the most studied unsupervised method for these features, and the results are compared. Furthermore, all experiments are repeated on noisy images. Results suggest that dimensionality reduction of these feature vectors can be accomplished successfully owing to the proposed method.

AB - Object recognition can be performed with high accuracy thanks to the robust feature descriptors defining the significant areas in images. However, these features suffer from high dimensional structure, in other words 'curse of dimensionality' for further processes. Autoencoders (AE) are proposed in this study to solve the dimensionality reduction problem of visual features. To assess the efficacy, object recognition is performed using reduced dimensional visual features. For this purpose, dimensionalities of three well-known feature vectors, namely, HOG, SIFT and SURF, are reduced to half. Moreover, deep learning based features are also reduced. Then, reduced vectors, which are called as AE-HOG, AE-SIFT, AE-SURF and AE-DEEP are fed to object recognition task. Also, dimensionality reduction is implemented by a variant of AE, variational autoencoder (VAE) and PCA, which is the most studied unsupervised method for these features, and the results are compared. Furthermore, all experiments are repeated on noisy images. Results suggest that dimensionality reduction of these feature vectors can be accomplished successfully owing to the proposed method.

KW - Autoencoder

KW - Dimensionality reduction

KW - HOG

KW - SIFT

KW - SURF

UR - http://www.scopus.com/inward/record.url?scp=85084857356&partnerID=8YFLogxK

U2 - 10.1109/SITIS.2019.00097

DO - 10.1109/SITIS.2019.00097

M3 - Conference contribution

AN - SCOPUS:85084857356

T3 - Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019

SP - 577

EP - 584

BT - Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019

A2 - Yetongnon, Kokou

A2 - Dipanda, Albert

A2 - Sanniti di Baja, Gabriella

A2 - Gallo, Luigi

A2 - Chbeir, Richard

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019

Y2 - 26 November 2019 through 29 November 2019

ER -

Keser RK , Toreyin BU. Autoencoder based dimensionality reduction of feature vectors for object recognition. In Yetongnon K, Dipanda A, Sanniti di Baja G, Gallo L, Chbeir R, editors, Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019. Institute of Electrical and Electronics Engineers Inc. 2019. p. 577-584. 9067860. (Proceedings - 15th International Conference on Signal Image Technology and Internet Based Systems, SISITS 2019). doi: 10.1109/SITIS.2019.00097

Autoencoder based dimensionality reduction of feature vectors for object recognition

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this