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 2017-12-17 10:30
Hierarchical Multi-stream STNs for Fine-grained Action Recognition
Dichao Liu, Yu Wang, Jien Kato (Nagoya Univ.) PRMU2017-103
Abstract (in Japanese) (See Japanese page) 
(in English) Fine-grained action recognition is a difficult problem mainly because that the discriminative information is usually very subtle.
Thus it is necessary to exploit multiple kinds (spatial and temporal) of information throughout different levels (from general to detail).
However, recent researches mainly works on the early aspect.

This paper pay special attention on the later aspect by proposing an approach to extract visual features in multiple hierarchies, namely general level, middle level and detailed level.
In order to obtain all-level spatial-temporal clues, we develop a 2-stream CNN model, with each stream consists of 3 sub-steam networks that correspond to the three levels.
Each sub-stream network consists a STN for localization and a CNN for feature extraction, which can be learned simultaneously.
The experiment results show that the proposed approach gain improvements over existing approaches.
Keyword (in Japanese) (See Japanese page) 
(in English) Action recognition / ConvNets / Fine-grained / Attention / / / /  
Reference Info. IEICE Tech. Rep., vol. 117, no. 362, PRMU2017-103, pp. 13-17, Dec. 2017.
Paper # PRMU2017-103 
Date of Issue 2017-12-10 (PRMU) 
ISSN Print edition: ISSN 0913-5685    Online edition: ISSN 2432-6380
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 PRMU2017-103

Conference Information
Committee PRMU  
Conference Date 2017-12-16 - 2017-12-17 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To PRMU 
Conference Code 2017-12-PRMU 
Language English (Japanese title is available) 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Hierarchical Multi-stream STNs for Fine-grained Action Recognition 
Sub Title (in English)  
Keyword(1) Action recognition  
Keyword(2) ConvNets  
Keyword(3) Fine-grained  
Keyword(4) Attention  
1st Author's Name Dichao Liu  
1st Author's Affiliation Nagoya University (Nagoya Univ.)
2nd Author's Name Yu Wang  
2nd Author's Affiliation Nagoya University (Nagoya Univ.)
3rd Author's Name Jien Kato  
3rd Author's Affiliation Nagoya University (Nagoya 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 Author-1 
Date Time 2017-12-17 10:30:00 
Presentation Time 30 minutes 
Registration for PRMU 
Paper # PRMU2017-103 
Volume (vol) vol.117 
Number (no) no.362 
Page pp.13-17 
Date of Issue 2017-12-10 (PRMU) 

[Return to Top Page]

[Return to IEICE Web Page]

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