Share Email Print
cover

Proceedings Paper

Image annotation by deep neural networks with attention shaping
Author(s): Kexin Zheng; Shaohe Lv; Fang Ma; Fei Chen; Chi Jin; Yong Dou
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Image annotation is a task of assigning semantic labels to an image. Recently, deep neural networks with visual attention have been utilized successfully in many computer vision tasks. In this paper, we show that conventional attention mechanism is easily misled by the salient class, i.e., the attended region always contains part of the image area describing the content of salient class at different attention iterations. To this end, we propose a novel attention shaping mechanism, which aims to maximize the non-overlapping area between consecutive attention processes by taking into account the history of previous attention vectors. Several weighting polices are studied to utilize the history information in different manners. In two benchmark datasets, i.e., PASCAL VOC2012 and MIRFlickr-25k, the average precision is improved by up to 10% in comparison with the state-of-the-art annotation methods.

Paper Details

Date Published: 21 July 2017
PDF: 7 pages
Proc. SPIE 10420, Ninth International Conference on Digital Image Processing (ICDIP 2017), 104201W (21 July 2017); doi: 10.1117/12.2281747
Show Author Affiliations
Kexin Zheng, National Univ. of Defense Technology (China)
Shaohe Lv, National Univ. of Defense Technology (China)
Fang Ma, National Univ. of Defense Technology (China)
Fei Chen, National Univ. of Defense Technology (China)
Chi Jin, Univ. of South China (China)
Yong Dou, National Univ. of Defense Technology (China)


Published in SPIE Proceedings Vol. 10420:
Ninth International Conference on Digital Image Processing (ICDIP 2017)
Charles M. Falco; Xudong Jiang, Editor(s)

© SPIE. Terms of Use
Back to Top