Title
EPAN: Effective Parts Attention Network for Scene Text Recognition
Abstract
•We propose the new EPAN framework for the flexible and robust recognition of scene text.•EPAN learns to decode characters of scene text from more informative vectors of the two dimensional feature map, which is more effective than previous work.•Trained with only word-level annotations, EPAN outperform state-of-the-art methods on several standard text recognition benchmarks.
Year
DOI
Venue
2020
10.1016/j.neucom.2019.10.010
Neurocomputing
Keywords
DocType
Volume
Scene text recognition,Optical character recognition,Deep neural network
Journal
376
ISSN
Citations 
PageRank 
0925-2312
1
0.35
References 
Authors
0
4
Name
Order
Citations
PageRank
Yunlong Huang110.35
Zenghui Sun2372.17
Lianwen Jin31337113.14
Canjie Luo4549.32