Dynamic multi-view hashing for online image retrieval

Liang Xie, Jialie Shen, Jungong Han, Lei Zhu, Ling Shao

Research output: Chapter in Book/Report/Conference proceedingConference Paper published in ProceedingsResearchpeer-review

Abstract

Advanced hashing technique is essential in large s-cale online image retrieval and organization, where image contents are frequently changed. While traditional multi-view hashing method has achieve promising effectiveness, its batch-based learning based scheme largely leads to very expensive updating cost. Meanwhile, existing online hashing scheme generally focuses on single-view data. Good effectiveness can not be expected when searching over real online images, which typically have multiple views. Further, both types of hashing methods only can generate hash codes with fixed length. Thus they have limited capability on comprehensive characterization of streaming image data. In this paper, we propose dynamic multi-view hashing (DMVH), which can adaptively augment hash codes according to dynamic changes of image. Meanwhile, DMVH leverages online learning to generate hash codes. It can increase the code length when current code is not able to represent new images effectively. Moreover, to gain further improvement on overall performance, each view is assigned with a weight, which can be efficiently updated in the online learning process. In order to avoid the frequent updating of code length and view weights, an intelligent buffering scheme is designed to preserve significant data to maintain good effectiveness of DMVH. Experimental results on t-wo real-world image datasets demonstrate superior performance of DWVH over several state-of-the-art hashing methods.

Original languageEnglish
Title of host publication26th International Joint Conference on Artificial Intelligence, IJCAI 2017
EditorsCarles Sierra
Place of PublicationMelbourne; Australia
PublisherInternational Joint Conferences on Artificial Intelligence
Pages3133-3139
Number of pages7
ISBN (Electronic)9780999241103
DOIs
Publication statusPublished - 1 Jan 2017
Externally publishedYes
Event26th International Joint Conference on Artificial Intelligence, IJCAI 2017 - Melbourne, Australia
Duration: 19 Aug 201725 Aug 2017

Conference

Conference26th International Joint Conference on Artificial Intelligence, IJCAI 2017
CountryAustralia
CityMelbourne
Period19/08/1725/08/17

Fingerprint

Image retrieval
Costs

Cite this

Xie, L., Shen, J., Han, J., Zhu, L., & Shao, L. (2017). Dynamic multi-view hashing for online image retrieval. In C. Sierra (Ed.), 26th International Joint Conference on Artificial Intelligence, IJCAI 2017 (pp. 3133-3139). Melbourne; Australia: International Joint Conferences on Artificial Intelligence. https://doi.org/10.24963/ijcai.2017/437
Xie, Liang ; Shen, Jialie ; Han, Jungong ; Zhu, Lei ; Shao, Ling. / Dynamic multi-view hashing for online image retrieval. 26th International Joint Conference on Artificial Intelligence, IJCAI 2017. editor / Carles Sierra. Melbourne; Australia : International Joint Conferences on Artificial Intelligence, 2017. pp. 3133-3139
@inproceedings{9f3a1b029b924fe2a58643392db04986,
title = "Dynamic multi-view hashing for online image retrieval",
abstract = "Advanced hashing technique is essential in large s-cale online image retrieval and organization, where image contents are frequently changed. While traditional multi-view hashing method has achieve promising effectiveness, its batch-based learning based scheme largely leads to very expensive updating cost. Meanwhile, existing online hashing scheme generally focuses on single-view data. Good effectiveness can not be expected when searching over real online images, which typically have multiple views. Further, both types of hashing methods only can generate hash codes with fixed length. Thus they have limited capability on comprehensive characterization of streaming image data. In this paper, we propose dynamic multi-view hashing (DMVH), which can adaptively augment hash codes according to dynamic changes of image. Meanwhile, DMVH leverages online learning to generate hash codes. It can increase the code length when current code is not able to represent new images effectively. Moreover, to gain further improvement on overall performance, each view is assigned with a weight, which can be efficiently updated in the online learning process. In order to avoid the frequent updating of code length and view weights, an intelligent buffering scheme is designed to preserve significant data to maintain good effectiveness of DMVH. Experimental results on t-wo real-world image datasets demonstrate superior performance of DWVH over several state-of-the-art hashing methods.",
author = "Liang Xie and Jialie Shen and Jungong Han and Lei Zhu and Ling Shao",
year = "2017",
month = "1",
day = "1",
doi = "10.24963/ijcai.2017/437",
language = "English",
pages = "3133--3139",
editor = "Carles Sierra",
booktitle = "26th International Joint Conference on Artificial Intelligence, IJCAI 2017",
publisher = "International Joint Conferences on Artificial Intelligence",

}

Xie, L, Shen, J, Han, J, Zhu, L & Shao, L 2017, Dynamic multi-view hashing for online image retrieval. in C Sierra (ed.), 26th International Joint Conference on Artificial Intelligence, IJCAI 2017. International Joint Conferences on Artificial Intelligence, Melbourne; Australia, pp. 3133-3139, 26th International Joint Conference on Artificial Intelligence, IJCAI 2017, Melbourne, Australia, 19/08/17. https://doi.org/10.24963/ijcai.2017/437

Dynamic multi-view hashing for online image retrieval. / Xie, Liang; Shen, Jialie; Han, Jungong; Zhu, Lei; Shao, Ling.

26th International Joint Conference on Artificial Intelligence, IJCAI 2017. ed. / Carles Sierra. Melbourne; Australia : International Joint Conferences on Artificial Intelligence, 2017. p. 3133-3139.

Research output: Chapter in Book/Report/Conference proceedingConference Paper published in ProceedingsResearchpeer-review

TY - GEN

T1 - Dynamic multi-view hashing for online image retrieval

AU - Xie, Liang

AU - Shen, Jialie

AU - Han, Jungong

AU - Zhu, Lei

AU - Shao, Ling

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Advanced hashing technique is essential in large s-cale online image retrieval and organization, where image contents are frequently changed. While traditional multi-view hashing method has achieve promising effectiveness, its batch-based learning based scheme largely leads to very expensive updating cost. Meanwhile, existing online hashing scheme generally focuses on single-view data. Good effectiveness can not be expected when searching over real online images, which typically have multiple views. Further, both types of hashing methods only can generate hash codes with fixed length. Thus they have limited capability on comprehensive characterization of streaming image data. In this paper, we propose dynamic multi-view hashing (DMVH), which can adaptively augment hash codes according to dynamic changes of image. Meanwhile, DMVH leverages online learning to generate hash codes. It can increase the code length when current code is not able to represent new images effectively. Moreover, to gain further improvement on overall performance, each view is assigned with a weight, which can be efficiently updated in the online learning process. In order to avoid the frequent updating of code length and view weights, an intelligent buffering scheme is designed to preserve significant data to maintain good effectiveness of DMVH. Experimental results on t-wo real-world image datasets demonstrate superior performance of DWVH over several state-of-the-art hashing methods.

AB - Advanced hashing technique is essential in large s-cale online image retrieval and organization, where image contents are frequently changed. While traditional multi-view hashing method has achieve promising effectiveness, its batch-based learning based scheme largely leads to very expensive updating cost. Meanwhile, existing online hashing scheme generally focuses on single-view data. Good effectiveness can not be expected when searching over real online images, which typically have multiple views. Further, both types of hashing methods only can generate hash codes with fixed length. Thus they have limited capability on comprehensive characterization of streaming image data. In this paper, we propose dynamic multi-view hashing (DMVH), which can adaptively augment hash codes according to dynamic changes of image. Meanwhile, DMVH leverages online learning to generate hash codes. It can increase the code length when current code is not able to represent new images effectively. Moreover, to gain further improvement on overall performance, each view is assigned with a weight, which can be efficiently updated in the online learning process. In order to avoid the frequent updating of code length and view weights, an intelligent buffering scheme is designed to preserve significant data to maintain good effectiveness of DMVH. Experimental results on t-wo real-world image datasets demonstrate superior performance of DWVH over several state-of-the-art hashing methods.

UR - http://www.scopus.com/inward/record.url?scp=85031930241&partnerID=8YFLogxK

U2 - 10.24963/ijcai.2017/437

DO - 10.24963/ijcai.2017/437

M3 - Conference Paper published in Proceedings

SP - 3133

EP - 3139

BT - 26th International Joint Conference on Artificial Intelligence, IJCAI 2017

A2 - Sierra, Carles

PB - International Joint Conferences on Artificial Intelligence

CY - Melbourne; Australia

ER -

Xie L, Shen J, Han J, Zhu L, Shao L. Dynamic multi-view hashing for online image retrieval. In Sierra C, editor, 26th International Joint Conference on Artificial Intelligence, IJCAI 2017. Melbourne; Australia: International Joint Conferences on Artificial Intelligence. 2017. p. 3133-3139 https://doi.org/10.24963/ijcai.2017/437