Scalable Hypergraph-Based Image Retrieval and Tagging System

Lu Chen, Yunjun Gao, Yuanliang Zhang, Sibo Wang, Baihua Zheng

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

Resumé

Massive amounts of images textually annotated by different users are provided by social image websites, e.g., Flickr. Social images are always associated with various information, such as visual features, tags, and users. In this paper, we utilize hypergraph instead of ordinary graph to model social images, since relations among various information are more sophisticated than pairwise. Based on the hypergraph, we propose HIRT, a scalable image retrieval and tagging system, which uses Personalized PageRank to measure vertex similarity, and employs top-k search to support image retrieval and tagging. To achieve good scalability and efficiency, we develop parallel and approximate top-k search algorithms with quality guarantees. Experiments on a large Flickr dataset confirm the effectiveness and efficiency of our proposed system HIRT compared with existing state-of-the-art hypergraph based image retrieval system. In addition, our parallel and approximate top-k search methods are verified to be more efficient than the state-of-the-art methods and meanwhile achieve higher result quality.
OriginalsprogEngelsk
TitelProceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018
Antal sider12
ForlagIEEE
Publikationsdato24 okt. 2018
Sider257-268
Artikelnummer8509253
ISBN (Trykt)978-1-5386-5521-4
ISBN (Elektronisk)978-1-5386-5520-7
DOI
StatusUdgivet - 24 okt. 2018
Begivenhed34th IEEE International Conference on Data Engineering, ICDE 2018 - Paris, Frankrig
Varighed: 16 apr. 201819 apr. 2018

Konference

Konference34th IEEE International Conference on Data Engineering, ICDE 2018
LandFrankrig
ByParis
Periode16/04/201819/04/2018

Fingerprint

Image retrieval
Scalability
Websites
Experiments

Citer dette

Chen, L., Gao, Y., Zhang, Y., Wang, S., & Zheng, B. (2018). Scalable Hypergraph-Based Image Retrieval and Tagging System. I Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018 (s. 257-268). [8509253] IEEE. https://doi.org/10.1109/ICDE.2018.00032
Chen, Lu ; Gao, Yunjun ; Zhang, Yuanliang ; Wang, Sibo ; Zheng, Baihua. / Scalable Hypergraph-Based Image Retrieval and Tagging System. Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018. IEEE, 2018. s. 257-268
@inproceedings{26169cb235204b9a8f10c9a188ae86c6,
title = "Scalable Hypergraph-Based Image Retrieval and Tagging System",
abstract = "Massive amounts of images textually annotated by different users are provided by social image websites, e.g., Flickr. Social images are always associated with various information, such as visual features, tags, and users. In this paper, we utilize hypergraph instead of ordinary graph to model social images, since relations among various information are more sophisticated than pairwise. Based on the hypergraph, we propose HIRT, a scalable image retrieval and tagging system, which uses Personalized PageRank to measure vertex similarity, and employs top-k search to support image retrieval and tagging. To achieve good scalability and efficiency, we develop parallel and approximate top-k search algorithms with quality guarantees. Experiments on a large Flickr dataset confirm the effectiveness and efficiency of our proposed system HIRT compared with existing state-of-the-art hypergraph based image retrieval system. In addition, our parallel and approximate top-k search methods are verified to be more efficient than the state-of-the-art methods and meanwhile achieve higher result quality.",
keywords = "Hypergraph, Image Retrieval, Image Tagging, System",
author = "Lu Chen and Yunjun Gao and Yuanliang Zhang and Sibo Wang and Baihua Zheng",
year = "2018",
month = "10",
day = "24",
doi = "10.1109/ICDE.2018.00032",
language = "English",
isbn = "978-1-5386-5521-4",
pages = "257--268",
booktitle = "Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018",
publisher = "IEEE",
address = "United States",

}

Chen, L, Gao, Y, Zhang, Y, Wang, S & Zheng, B 2018, Scalable Hypergraph-Based Image Retrieval and Tagging System. i Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018., 8509253, IEEE, s. 257-268, 34th IEEE International Conference on Data Engineering, ICDE 2018, Paris, Frankrig, 16/04/2018. https://doi.org/10.1109/ICDE.2018.00032

Scalable Hypergraph-Based Image Retrieval and Tagging System. / Chen, Lu; Gao, Yunjun; Zhang, Yuanliang; Wang, Sibo; Zheng, Baihua.

Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018. IEEE, 2018. s. 257-268 8509253.

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

TY - GEN

T1 - Scalable Hypergraph-Based Image Retrieval and Tagging System

AU - Chen, Lu

AU - Gao, Yunjun

AU - Zhang, Yuanliang

AU - Wang, Sibo

AU - Zheng, Baihua

PY - 2018/10/24

Y1 - 2018/10/24

N2 - Massive amounts of images textually annotated by different users are provided by social image websites, e.g., Flickr. Social images are always associated with various information, such as visual features, tags, and users. In this paper, we utilize hypergraph instead of ordinary graph to model social images, since relations among various information are more sophisticated than pairwise. Based on the hypergraph, we propose HIRT, a scalable image retrieval and tagging system, which uses Personalized PageRank to measure vertex similarity, and employs top-k search to support image retrieval and tagging. To achieve good scalability and efficiency, we develop parallel and approximate top-k search algorithms with quality guarantees. Experiments on a large Flickr dataset confirm the effectiveness and efficiency of our proposed system HIRT compared with existing state-of-the-art hypergraph based image retrieval system. In addition, our parallel and approximate top-k search methods are verified to be more efficient than the state-of-the-art methods and meanwhile achieve higher result quality.

AB - Massive amounts of images textually annotated by different users are provided by social image websites, e.g., Flickr. Social images are always associated with various information, such as visual features, tags, and users. In this paper, we utilize hypergraph instead of ordinary graph to model social images, since relations among various information are more sophisticated than pairwise. Based on the hypergraph, we propose HIRT, a scalable image retrieval and tagging system, which uses Personalized PageRank to measure vertex similarity, and employs top-k search to support image retrieval and tagging. To achieve good scalability and efficiency, we develop parallel and approximate top-k search algorithms with quality guarantees. Experiments on a large Flickr dataset confirm the effectiveness and efficiency of our proposed system HIRT compared with existing state-of-the-art hypergraph based image retrieval system. In addition, our parallel and approximate top-k search methods are verified to be more efficient than the state-of-the-art methods and meanwhile achieve higher result quality.

KW - Hypergraph

KW - Image Retrieval

KW - Image Tagging

KW - System

UR - http://www.scopus.com/inward/record.url?scp=85057109375&partnerID=8YFLogxK

U2 - 10.1109/ICDE.2018.00032

DO - 10.1109/ICDE.2018.00032

M3 - Article in proceeding

SN - 978-1-5386-5521-4

SP - 257

EP - 268

BT - Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018

PB - IEEE

ER -

Chen L, Gao Y, Zhang Y, Wang S, Zheng B. Scalable Hypergraph-Based Image Retrieval and Tagging System. I Proceedings - IEEE 34th International Conference on Data Engineering, ICDE 2018. IEEE. 2018. s. 257-268. 8509253 https://doi.org/10.1109/ICDE.2018.00032