Tensor Decomposition of Dense SIFT Descriptors in Object Recognition

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

Abstract

In machine vision, Scale-invariant feature transform (SIFT) and its variants have been widely used in image classification task. However, the high dimensionality nature of SIFT features, usually in the order of multiple thousands per image, would require careful consideration in place to achieve accurate and timely categorization of objects within images. This paper explores the possibility of processing SIFT features as tensors and uses tensor decomposition techniques on high-order SIFT tensors for dimensionality reduction. The method focuses on both accuracy and efficiency aspects and the validation result with the Caltech 101 dataset confirms the improvement with notable margins.

Original languageEnglish
Title of host publicationESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning
EditorsMichel Verleysen
Place of PublicationBruges, Belgium
PublisherEuropean Symposium of Artificial Neural Networks
Pages319-324
Number of pages6
Volume1
ISBN (Print)9782874190957
Publication statusPublished - 23 Apr 2014
EventEuropean Symposium on Artificial Neural Networks: ESANN 2014 - Brugge, Brugge, Belgium
Duration: 23 Apr 201425 Apr 2014

Publication series

Name22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings

Conference

ConferenceEuropean Symposium on Artificial Neural Networks
Abbreviated titleESANN 2014
CountryBelgium
CityBrugge
Period23/04/1425/04/14

Fingerprint

Object recognition
Tensors
Decomposition
Image classification
Computer vision
Processing

Cite this

Vo, T., TRAN, D., & MA, W. (2014). Tensor Decomposition of Dense SIFT Descriptors in Object Recognition. In M. Verleysen (Ed.), ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning (Vol. 1, pp. 319-324). (22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings). Bruges, Belgium: European Symposium of Artificial Neural Networks.
Vo, Tan ; TRAN, Dat ; MA, Wanli. / Tensor Decomposition of Dense SIFT Descriptors in Object Recognition. ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning. editor / Michel Verleysen. Vol. 1 Bruges, Belgium : European Symposium of Artificial Neural Networks, 2014. pp. 319-324 (22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings).
@inproceedings{fad736c30ae04f50b54af243d0fcf901,
title = "Tensor Decomposition of Dense SIFT Descriptors in Object Recognition",
abstract = "In machine vision, Scale-invariant feature transform (SIFT) and its variants have been widely used in image classification task. However, the high dimensionality nature of SIFT features, usually in the order of multiple thousands per image, would require careful consideration in place to achieve accurate and timely categorization of objects within images. This paper explores the possibility of processing SIFT features as tensors and uses tensor decomposition techniques on high-order SIFT tensors for dimensionality reduction. The method focuses on both accuracy and efficiency aspects and the validation result with the Caltech 101 dataset confirms the improvement with notable margins.",
keywords = "Tensor-Decomposition, SIFT Descriptors, Object Recognition",
author = "Tan Vo and Dat TRAN and Wanli MA",
year = "2014",
month = "4",
day = "23",
language = "English",
isbn = "9782874190957",
volume = "1",
series = "22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings",
publisher = "European Symposium of Artificial Neural Networks",
pages = "319--324",
editor = "Michel Verleysen",
booktitle = "ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning",

}

Vo, T, TRAN, D & MA, W 2014, Tensor Decomposition of Dense SIFT Descriptors in Object Recognition. in M Verleysen (ed.), ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning. vol. 1, 22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings, European Symposium of Artificial Neural Networks, Bruges, Belgium, pp. 319-324, European Symposium on Artificial Neural Networks, Brugge, Belgium, 23/04/14.

Tensor Decomposition of Dense SIFT Descriptors in Object Recognition. / Vo, Tan; TRAN, Dat; MA, Wanli.

ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning. ed. / Michel Verleysen. Vol. 1 Bruges, Belgium : European Symposium of Artificial Neural Networks, 2014. p. 319-324 (22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings).

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

TY - GEN

T1 - Tensor Decomposition of Dense SIFT Descriptors in Object Recognition

AU - Vo, Tan

AU - TRAN, Dat

AU - MA, Wanli

PY - 2014/4/23

Y1 - 2014/4/23

N2 - In machine vision, Scale-invariant feature transform (SIFT) and its variants have been widely used in image classification task. However, the high dimensionality nature of SIFT features, usually in the order of multiple thousands per image, would require careful consideration in place to achieve accurate and timely categorization of objects within images. This paper explores the possibility of processing SIFT features as tensors and uses tensor decomposition techniques on high-order SIFT tensors for dimensionality reduction. The method focuses on both accuracy and efficiency aspects and the validation result with the Caltech 101 dataset confirms the improvement with notable margins.

AB - In machine vision, Scale-invariant feature transform (SIFT) and its variants have been widely used in image classification task. However, the high dimensionality nature of SIFT features, usually in the order of multiple thousands per image, would require careful consideration in place to achieve accurate and timely categorization of objects within images. This paper explores the possibility of processing SIFT features as tensors and uses tensor decomposition techniques on high-order SIFT tensors for dimensionality reduction. The method focuses on both accuracy and efficiency aspects and the validation result with the Caltech 101 dataset confirms the improvement with notable margins.

KW - Tensor-Decomposition

KW - SIFT Descriptors

KW - Object Recognition

UR - http://technodocbox.com/3D_Graphics/66588124-Tensor-decomposition-of-dense-sift-descriptors-in-object-recognition.html

UR - http://www.scopus.com/inward/record.url?scp=84961990539&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9782874190957

VL - 1

T3 - 22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings

SP - 319

EP - 324

BT - ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning

A2 - Verleysen, Michel

PB - European Symposium of Artificial Neural Networks

CY - Bruges, Belgium

ER -

Vo T, TRAN D, MA W. Tensor Decomposition of Dense SIFT Descriptors in Object Recognition. In Verleysen M, editor, ESANN 2014 European Symposium of Artificial Neural Networks, Computational Intelligence and Machine Learning. Vol. 1. Bruges, Belgium: European Symposium of Artificial Neural Networks. 2014. p. 319-324. (22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2014 - Proceedings).