Image retrieval based on ResNet and ITQ

DOI

10.2991/mecae-18.2018.96How to use a DOI?

Keywords

Learning to Hash, Deep Residual Network, Iterative Quantization.

Abstract

In recent years, more and more hash learning methods have been applied to solve large-scale vision problems. It has been shown that learning hash function by using supervised information can boost hashing quality. The state-of-the-art image retrieval hashing methods based on visual features lacks of learning ability, the image expression ability is weak and the efficiency of large-scale image retrieval is low. In this paper, we propose a new supervised hashing framework based on deep Residual Networks and Iterative Quantization hashing. Firstly, we exploit the learning abilities of deep residual network to mine the inherent hidden relationship of image content, extract deep feature descriptors, and increase the visual expression of images Secondly, Iterative Quantization Hashing is applied to learn from the high-dimensional image feature and map into low-dimensional hamming space and achieve compact Hash codes. Finally, image retrieval is accomplished in low-dimensional hamming space. Experimental results of MNIST, CIFAR-10, CIFAR-100 and Caltech 256 show that the expression ability of visual feature is effectively improved and the image retrieval performance is substantially boosted compared with other related methods.

Copyright

© 2018, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Guijun Wang
AU  - Baohua Qiang
AU  - Xianchun Zou
AU  - Jinyun Lu
PY  - 2018/03
DA  - 2018/03
TI  - Image retrieval based on ResNet and ITQ
BT  - Proceedings of the 2018 International Conference on Mechanical, Electronic, Control and Automation Engineering (MECAE 2018)
PB  - Atlantis Press
SP  - 491
EP  - 496
SN  - 2352-5401
UR  - https://doi.org/10.2991/mecae-18.2018.96
DO  - 10.2991/mecae-18.2018.96
ID  - Wang2018/03
ER  -