Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2016
Session Number:T1-5
Session:
Number:4507
Scene Text Localization Using A Two-Class Detector Constructed By Filtered Channel Features
Takuro Oki, Ryusuke Miyamoto ,
pp.459-462
Publication Date:2016/7/10
Online ISSN:2188-5079
DOI:10.34385/proc.61.4507
PDF download (3.2MB)
Summary:
The authors have tried to improve both accuracy and processing speed for scene text localization using integral channel features for construction of detector in the previous work but sufficient accuracy did not achieved. To improve the detection accuracy of the framework, this paper tries to apply filtered channel features that shows excellent and much better performance than deep learning for human detection.