IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2021-10-09 10:45
Moving Scene Text Detection Using Synthetic Scene Text Video for Training
Zhiyuan Xie, Hideaki Goto, Takuo Suganuma (Tohoku Univ.) PRMU2021-21
Abstract (in Japanese) (See Japanese page) 
(in English) In computer vision areas, scene text is valuable information for applications including scene understanding, autopilot, blind assistance, etc. Such applications usually work with moving camera and analyze information from video. Benefiting from its self-encoding feature, a neural network is capable of learning pattern from existing datasets and it is widely used at image scene text processing areas. Subject to the amount and quality of video scene text datasets, research in video scene text processing area usually uses image-based neural network or pure image processing algorithm to analyze the information. To overcome this limitation, we have developed a Synthetic Scene Text Video (SSTV) method. It is a data augmentation method which can generate synthetic video from single labeled image based on a synthetic trajectory. Additionally, we propose a magnitude-based mask which alleviates the ambiguity in texts with complex shape and improves the performance compared with conventional binary mask. Using this magnitude-based mask and the SSTV, the F-score of video scene text detection task is improved from 46.34% to 83.47% in comparison with conventional image-based detection network.
Keyword (in Japanese) (See Japanese page) 
(in English) Data augmentation / video scene text detection / Gaussian intensity mask / / / / /  
Reference Info. IEICE Tech. Rep., vol. 121, no. 192, PRMU2021-21, pp. 28-33, Oct. 2021.
Paper # PRMU2021-21 
Date of Issue 2021-10-01 (PRMU) 
ISSN Print edition: ISSN 0913-5685  Online edition: ISSN 2432-6380
Copyright
and
reproduction
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF PRMU2021-21

Conference Information
Committee PRMU  
Conference Date 2021-10-08 - 2021-10-09 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Processes and technologies to make research more efficient 
Paper Information
Registration To PRMU 
Conference Code 2021-10-PRMU 
Language English 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Moving Scene Text Detection Using Synthetic Scene Text Video for Training 
Sub Title (in English)  
Keyword(1) Data augmentation  
Keyword(2) video scene text detection  
Keyword(3) Gaussian intensity mask  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Zhiyuan Xie  
1st Author's Affiliation Tohoku University (Tohoku Univ.)
2nd Author's Name Hideaki Goto  
2nd Author's Affiliation Tohoku University (Tohoku Univ.)
3rd Author's Name Takuo Suganuma  
3rd Author's Affiliation Tohoku University (Tohoku Univ.)
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker
Date Time 2021-10-09 10:45:00 
Presentation Time 15 
Registration for PRMU 
Paper # IEICE-PRMU2021-21 
Volume (vol) IEICE-121 
Number (no) no.192 
Page pp.28-33 
#Pages IEICE-6 
Date of Issue IEICE-PRMU-2021-10-01 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan