Efficient Image Search with Neural Net Features

D 2015

Efficient Image Search with Neural Net Features

NOVÁK, David, Jan ČECH and Pavel ZEZULA

Basic information

Original name

Efficient Image Search with Neural Net Features

Authors

NOVÁK, David (203 Czech Republic, guarantor, belonging to the institution), Jan ČECH (203 Czech Republic) and Pavel ZEZULA (203 Czech Republic, belonging to the institution)

Edition

New York, Similarity Search and Applications: 8th International Conference, SISAP 2015, Glasgow, UK, October 12-14, 2015, Proceedings, p. 237-243, 7 pp. 2015

Publisher

Springer International Publishing

Other information

Language

English

Type of outcome

Stať ve sborníku

Field of Study

10201 Computer sciences, information science, bioinformatics

Country of publisher

United States of America

Confidentiality degree

není předmětem státního či obchodního tajemství

Publication form

printed version "print"

References:

DOI link

Impact factor

Impact factor: 0.402 in 2005

RIV identification code

RIV/00216224:14330/15:00081692

Organization unit

Faculty of Informatics

ISBN

978-3-319-25086-1

ISSN

DOI

http://dx.doi.org/10.1007/978-3-319-25087-8_22

UT WoS

000374289600022

Keywords in English

metric indexing; deep convolutional neural network; contentbased image retrieval

Abstract

V originále

We present an efficiency evaluation of similarity search techniques applied on visual features from deep neural networks. Our test collection consists of 20 million 4096-dimensional descriptors (320GB of data). We test approximate k-NN search using several techniques, specifically FLANN library (a popular in-memory implementation of k-d tree forest), M-Index (that uses recursive Voronoi partitioning of a metric space), and PPP-Codes, which work with memory codes of metric objects and use disk storage for candidate refinement. Our evaluation shows that as long as the data fit in main memory, the FLANN and the M-Index have practically the same ratio between precision and response time. The PPP-Codes identify candidate sets ten times smaller then the other techniques and the response times are around 500 ms for the whole 20M dataset stored on the disk. The visual search with this index is available as an online demo application. The collection of 20M descriptors is provided as a public dataset to academic community.

Links

GAP103/10/0886, research and development project

Name: Vizuální vyhledávání obrázků na Webu (Acronym: VisualWeb)

Investor: Czech Science Foundation, Content-based Image Retrieval on the Web Scale

Citovat

NOVÁK, David, Jan ČECH and Pavel ZEZULA. Efficient Image Search with Neural Net Features. In Similarity Search and Applications: 8th International Conference, SISAP 2015, Glasgow, UK, October 12-14, 2015, Proceedings. New York: Springer International Publishing, 2015, p. 237-243. ISBN 978-3-319-25086-1. Available from: https://dx.doi.org/10.1007/978-3-319-25087-8_22.

@inproceedings{1341689,
   author = {Novák, David and Čech, Jan and Zezula, Pavel},
   address = {New York},
   booktitle = {Similarity Search and Applications: 8th International Conference, SISAP 2015, Glasgow, UK, October 12-14, 2015, Proceedings},
   doi = {http://dx.doi.org/10.1007/978-3-319-25087-8_22},
   keywords = {metric indexing; deep convolutional neural network; contentbased image retrieval},
   howpublished = {tištěná verze "print"},
   language = {eng},
   location = {New York},
   isbn = {978-3-319-25086-1},
   pages = {237-243},
   publisher = {Springer International Publishing},
   title = {Efficient Image Search with Neural Net Features},
   url = {http://dx.doi.org/10.1007/978-3-319-25087-8_22},
   year = {2015}
}

TY  - JOUR
ID  - 1341689
AU  - Novák, David - Čech, Jan - Zezula, Pavel
PY  - 2015
TI  - Efficient Image Search with Neural Net Features
PB  - Springer International Publishing
CY  - New York
SN  - 9783319250861
KW  - metric indexing
KW  - deep convolutional neural network
KW  - contentbased image retrieval
UR  - http://dx.doi.org/10.1007/978-3-319-25087-8_22
L2  - http://dx.doi.org/10.1007/978-3-319-25087-8_22
N2  - We present an efficiency evaluation of similarity search techniques applied on visual features from deep neural networks. Our test collection consists of 20 million 4096-dimensional descriptors (320GB of data). We test approximate k-NN search using several techniques, specifically FLANN library (a popular in-memory implementation of k-d tree forest), M-Index (that uses recursive Voronoi partitioning of a metric space), and PPP-Codes, which work with memory codes of metric objects and use disk storage for candidate refinement. Our evaluation shows that as long as the data fit in main memory, the FLANN and the M-Index have practically the same ratio between precision and response time. The PPP-Codes identify candidate sets ten times smaller then the other techniques and the response times are around 500 ms for the whole 20M dataset stored on the disk. The visual search with this index is available as an online demo application. The collection of 20M descriptors is provided as a public dataset to academic community.
ER  -

NOVÁK, David, Jan ČECH and Pavel ZEZULA. Efficient Image Search with Neural Net Features. In \textit{Similarity Search and Applications: 8th International Conference, SISAP 2015, Glasgow, UK, October 12-14, 2015, Proceedings}. New York: Springer International Publishing, 2015, p.~237-243. ISBN~978-3-319-25086-1. Available from: https://dx.doi.org/10.1007/978-3-319-25087-8\_{}22.

Detailed Information on Publication Record