| 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 | 2017-03-21 15:25 Aggregating Appearance and Motion Information using LSTM for Moving Object Detection Tuan Tu Trinh, Ryota Yoshihashi, Rei Kawakami (Todai), Shaodi You (Data61-CSIRO, ANU), Makoto Iida, Takeshi Naemura (Todai)  BioX2016-70 PRMU2016-233 |
|
| Abstract | (in Japanese) | (See Japanese page) |
| (in English) | Recently, Convolutional Neural Networks (CNNs) have shown impressive results in still image data for the reason that they can extract more data-driven features compared to traditional manually designed features. However, in real-life problems, objects may have very low resolution and hardly recognized or distinguished from hard-negatives by human eyes. In fact, target object could be easily detected if motion is considered, especially when detecting very small objects in large scene. Several studies have followed the idea and shown that detection performance can be improved by combining motion features with static ones. However, how to utilize motion appropriately to achieve the best performance in detection is still in debate. Most of previous studies incorporate motion only through handcrafted features and the main ideas are removing background and keeping contour of moving objects. However, it is difficult to design feature that can represent motion in data-sets where objects are captured in very low-resolution. In this study, we proposed a pipeline combining CNNs and LSTM which are capable of learning long-term dependencies from continuous information for object detection, exploiting as much information as possible from low-resolution input. | |
| Keyword | (in Japanese) | (See Japanese page) |
| (in English) | Deep learning / CNN / LSTM / Motion / Detection / Moving object / Low resolution / | |
| Reference Info. | IEICE Tech. Rep., vol. 116, no. 528, PRMU2016-233, pp. 221-226, March 2017. | |
| Paper # | PRMU2016-233 | |
| Date of Issue | 2017-03-13 (BioX, 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 | BioX2016-70 PRMU2016-233 |
|
| Conference Information | |
| Committee | PRMU BioX |
| Conference Date | 2017-03-20 - 2017-03-21 |
| 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 | 2017-03-PRMU-BioX |
| Language | Japanese |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Aggregating Appearance and Motion Information using LSTM for Moving Object Detection |
| Sub Title (in English) | |
| Keyword(1) | Deep learning |
| Keyword(2) | CNN |
| Keyword(3) | LSTM |
| Keyword(4) | Motion |
| Keyword(5) | Detection |
| Keyword(6) | Moving object |
| Keyword(7) | Low resolution |
| Keyword(8) | |
| 1st Author's Name | Tuan Tu Trinh |
| 1st Author's Affiliation | The University of Tokyo (Todai) |
| 2nd Author's Name | Ryota Yoshihashi |
| 2nd Author's Affiliation | The University of Tokyo (Todai) |
| 3rd Author's Name | Rei Kawakami |
| 3rd Author's Affiliation | The University of Tokyo (Todai) |
| 4th Author's Name | Shaodi You |
| 4th Author's Affiliation | Data61-CSIRO, Australian National University (Data61-CSIRO, ANU) |
| 5th Author's Name | Makoto Iida |
| 5th Author's Affiliation | The University of Tokyo (Todai) |
| 6th Author's Name | Takeshi Naemura |
| 6th Author's Affiliation | The University of Tokyo (Todai) |
| Speaker | Author-1 |
| Date Time | 2017-03-21 15:25:00 |
| Presentation Time | 25 minutes |
| Registration for | PRMU |
| Paper # | BioX2016-70, PRMU2016-233 |
| Volume (vol) | vol.116 |
| Number (no) | no.527(BioX), no.528(PRMU) |
| Page | pp.221-226 |
| #Pages | 6 |
| Date of Issue | 2017-03-13 (BioX, PRMU) |
[Return to Top Page]