| 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 | |
| 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 | 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 |
| Keyword(5) | |
| Keyword(6) | |
| Keyword(7) | |
| Keyword(8) | |
| 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 | () |
| 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 |
| #Pages | 5 |
| Date of Issue | 2017-12-10 (PRMU) |
[Return to Top Page]