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 2022-05-13 10:30
Visualization of Decision Rationale Using Social and Physical Attention Mechanisms in Human Trajectory Prediction Model
Masahiro Kato, Norimichi Ukita (TTI) PRMU2022-3
Abstract (in Japanese) (See Japanese page) 
(in English) There is a great deal of interest in explainable AI that clarifies the basis of decisions, such as why a model makes a prediction, for applications of computer vision technology in medicine, surveillance systems, automatic driving, and so on. However, visualization research of predictive and generative models is less explored than that of CNN models. In this study, we will take the lead in this unexplored field by researching the visualization of the basis for decision making in trajectory prediction models, a type of action prediction model. This study deals with a trajectory prediction model with two types of attention mechanisms: a physical attention mechanism that captures the interrelationships between people and space, and a social attention mechanism that captures the interrelationships between people. The physical attention mechanism visualizes the spatial region that influenced the prediction in static spatial information from an overhead image. The social attention mechanism visualizes the person who influenced the prediction based on the interrelationship between people. From each visualization, it is possible to interpret which parts of the surrounding spatial information the trajectory prediction model considered important, and which people it considered important in predicting the trajectory of a person. Experiments have shown reasonable prediction results without taking into account key areas and key people, which would be intuitively correct from a human perspective. Therefore, conventional trajectory prediction models may not handle the expected information well.
Keyword (in Japanese) (See Japanese page) 
(in English) Automatic Vehicle / Explainable AI / Action Prediction / Visualization / Attention Mechanism / Trajectory Prediction / /  
Reference Info. IEICE Tech. Rep., vol. 122, no. 13, PRMU2022-3, pp. 12-17, May 2022.
Paper # PRMU2022-3 
Date of Issue 2022-05-05 (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 PRMU2022-3

Conference Information
Committee PRMU IPSJ-CVIM  
Conference Date 2022-05-12 - 2022-05-13 
Place (in Japanese) (See Japanese page) 
Place (in English) Toyota Technological Institute 
Topics (in Japanese) (See Japanese page) 
Topics (in English) How to conduct research (post-graduation project for students) 
Paper Information
Registration To PRMU 
Conference Code 2022-05-PRMU-CVIM 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Visualization of Decision Rationale Using Social and Physical Attention Mechanisms in Human Trajectory Prediction Model 
Sub Title (in English)  
Keyword(1) Automatic Vehicle  
Keyword(2) Explainable AI  
Keyword(3) Action Prediction  
Keyword(4) Visualization  
Keyword(5) Attention Mechanism  
Keyword(6) Trajectory Prediction  
Keyword(7)  
Keyword(8)  
1st Author's Name Masahiro Kato  
1st Author's Affiliation Toyota Technological Institute (TTI)
2nd Author's Name Norimichi Ukita  
2nd Author's Affiliation Toyota Technological Institute (TTI)
3rd Author's Name  
3rd Author's Affiliation ()
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 2022-05-13 10:30:00 
Presentation Time 15 minutes 
Registration for PRMU 
Paper # PRMU2022-3 
Volume (vol) vol.122 
Number (no) no.13 
Page pp.12-17 
#Pages
Date of Issue 2022-05-05 (PRMU) 


[Return to Top Page]

[Return to IEICE Web Page]


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