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 2021-12-17 10:45
Data augmentation for domain adaptation by class consistency using soft target
Wataru Imaeda (Kyocera Corp.), Tsubasa Hirakawa, Takayoshi Yamashita, Hironobu Fujiyoshi (Chubu Univ.) PRMU2021-49
Abstract (in Japanese) (See Japanese page) 
(in English) Domain adaptation is a task to acquire sufficient discrimination performance between images with different domains in learning and evaluation. In this task, there are approaches such as acquisition of domain invariant feature quantity and elimination of domain gap using style transformation. In the approach by the style transformation, the input/output relation in the style transformation is 1: 1 in the conventional domain adaptation technique. Therefore, the transformed image with changed class information does not contribute to the improvement of the discrimination performance. In this paper, we propose a data augmentation method for domain adaptation using style transformation considering class information. In the proposed method, the transformation suitable for domain adaptation is realized by considering class consistency in image transformation using Soft target. In addition, by applying Generative Adversarial Networks, the style augmentation which generates multiple images with different styles from one image is carried out. We also show that the style augmentation is effective in identification using the Test-time Augmentation.
Keyword (in Japanese) (See Japanese page) 
(in English) Style transfer / Domain adaptation / Test-time Augmentation / GANs / / / /  
Reference Info. IEICE Tech. Rep., vol. 121, no. 304, PRMU2021-49, pp. 136-141, Dec. 2021.
Paper # PRMU2021-49 
Date of Issue 2021-12-09 (PRMU) 
ISSN 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 PRMU2021-49

Conference Information
Committee PRMU  
Conference Date 2021-12-16 - 2021-12-17 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To PRMU 
Conference Code 2021-12-PRMU 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Data augmentation for domain adaptation by class consistency using soft target 
Sub Title (in English)  
Keyword(1) Style transfer  
Keyword(2) Domain adaptation  
Keyword(3) Test-time Augmentation  
Keyword(4) GANs  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Wataru Imaeda  
1st Author's Affiliation Kyocera Corporation (Kyocera Corp.)
2nd Author's Name Tsubasa Hirakawa  
2nd Author's Affiliation Chubu University (Chubu Univ.)
3rd Author's Name Takayoshi Yamashita  
3rd Author's Affiliation Chubu University (Chubu Univ.)
4th Author's Name Hironobu Fujiyoshi  
4th Author's Affiliation Chubu University (Chubu Univ.)
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 Author-1 
Date Time 2021-12-17 10:45:00 
Presentation Time 15 minutes 
Registration for PRMU 
Paper # PRMU2021-49 
Volume (vol) vol.121 
Number (no) no.304 
Page pp.136-141 
#Pages
Date of Issue 2021-12-09 (PRMU) 


[Return to Top Page]

[Return to IEICE Web Page]


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