A Character Flow Framework for Multi-Oriented Scene Text Detection

作     者：Wen-Jun Yang Bei-Ji Zou Kai-Wen Li Shu Liu Wen-Jun Yang;Bei-Ji Zou;Kai-Wen Li;Shu Liu

作者机构：School of Computer Science and EngineeringCentral South UniversityChangsha410083China Hunan Engineering Research Center of Machine Vision and Intelligent MedicineChangsha410083China 

出 版 物：《Journal of Computer Science & Technology》 (计算机科学技术学报（英文版）)

年 卷 期：2021年第36卷第3期

页      面：465-477页

核心收录：

学科分类：081203[工学-计算机应用技术] 08[工学] 0835[工学-软件工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the National Natural Science Foundation of China under Grant No.61902435 the National Science and Technology Major Project of China under Grant No.2018AAA0102102 the 111 Project under Grant No.B18059 the Hunan Provincial Natural Science Foundation of China under Grant No.2019JJ50808 

主　　题：multi-oriented scene text detection character instance segmentation character flow feature pyramid network(FPN) bidirectional long short-term memory(BLSTM) 

摘      要：Scene text detection plays a significant role in various applications,such as object recognition,document management,and visual *** instance segmentation based method has been mostly used in existing research due to its advantages in dealing with multi-oriented ***,a large number of non-text pixels exist in the labels during the model training,leading to text *** this paper,we propose a novel multi-oriented scene text detection framework,which includes two main modules:character instance segmentation(one instance corresponds to one character),and character flow construction(one character flow corresponds to one word).We use feature pyramid network(FPN)to predict character and non-character instances with arbitrary directions.A joint network of FPN and bidirectional long short-term memory(BLSTM)is developed to explore the context information among isolated characters,which are finally grouped into character *** experiments are conducted on ICDAR2013,ICDAR2015,MSRA-TD500 and MLT datasets to demonstrate the effectiveness of our *** F-measures are 92.62%,88.02%,83.69%and 77.81%,respectively.