Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
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 2026-02-20 10:45
A PEFT Approach for SAM Considering Autonomous-Driving Class Labels
Yuki Masuko, Joonho Lee, Ryosuke Kawata, Shunsuke Kamijo (UTokyo) ITS2025-66 IE2025-81
Abstract (in Japanese) (See Japanese page) 
(in English) The existing method EVF-SAM (Early Vision–Language Fusion with Segment Anything Model) demonstrates high generality by combining the strong shape recognition capability of SAM with the language–image understanding ability of BEiT-3. However, it has limitations in fully reproducing the class labels specific to a given dataset. In particular, discrepancies in detection ranges for single instances were observed, making appropriate performance evaluation difficult. Moreover, segmentation for autonomous driving requires extensive labeling efforts for each class, which poses a major obstacle to practical deployment. Although rapid advances in computer vision have significantly improved image recognition accuracy, autonomous driving systems demand real-time performance under limited computational resources. Therefore, model designs that take computational cost into account are essential for practical use. In this study, to address these technical and practical challenges, we investigate an approach that introduces class-specific adapters, which learn knowledge independently for each class, into EVF-SAM. Specifically, we propose a segmentation model design that can correct class boundary errors through low-cost training and improve IoU (Intersection over Union) accuracy. Through this approach, appropriate performance evaluation and improved recognition accuracy can be achieved by correcting class label boundaries, while enabling efficient learning and dataset construction for autonomous driving. As a result, this study is expected to contribute to the realization of more reliable autonomous driving systems.
Keyword (in Japanese) (See Japanese page) 
(in English) Segmentation / Labeling / Computational Cost / EVF-SAM / Adapter / LoRA / /  
Reference Info. IEICE Tech. Rep., vol. 125, no. 355, ITS2025-66, pp. 153-158, Feb. 2026.
Paper # ITS2025-66 
Date of Issue 2026-02-12 (ITS, IE) 
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 ITS2025-66 IE2025-81

Conference Information
Committee IE ITS ITE-MMS ITE-ME ITE-AIT ITE-SIP  
Conference Date 2026-02-19 - 2026-02-20 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To ITS 
Conference Code 2026-02-IE-ITS-MMS-ME-AIT-SIP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) A PEFT Approach for SAM Considering Autonomous-Driving Class Labels 
Sub Title (in English)  
Keyword(1) Segmentation  
Keyword(2) Labeling  
Keyword(3) Computational Cost  
Keyword(4) EVF-SAM  
Keyword(5) Adapter  
Keyword(6) LoRA  
Keyword(7)  
Keyword(8)  
1st Author's Name Yuki Masuko  
1st Author's Affiliation The University of Tokyo (UTokyo)
2nd Author's Name Joonho Lee  
2nd Author's Affiliation The University of Tokyo (UTokyo)
3rd Author's Name Ryosuke Kawata  
3rd Author's Affiliation The University of Tokyo (UTokyo)
4th Author's Name Shunsuke Kamijo  
4th Author's Affiliation The University of Tokyo (UTokyo)
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 ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2026-02-20 10:45:00 
Presentation Time 15 minutes 
Registration for ITS 
Paper # ITS2025-66, IE2025-81 
Volume (vol) vol.125 
Number (no) no.355(ITS), no.356(IE) 
Page pp.153-158 
#Pages 6 
Date of Issue 2026-02-12 (ITS, IE) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan