Distribution-aware contrastive learning for domain adaptation in 3D LiDAR segmentation

IF 4.3 3区 计算机科学 Q2 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
Lamiae El Mendili, Sylvie Daniel, Thierry Badard
{"title":"Distribution-aware contrastive learning for domain adaptation in 3D LiDAR segmentation","authors":"Lamiae El Mendili,&nbsp;Sylvie Daniel,&nbsp;Thierry Badard","doi":"10.1016/j.cviu.2025.104438","DOIUrl":null,"url":null,"abstract":"<div><div>Semantic segmentation of 3D LiDAR point clouds is very important for applications like autonomous driving and digital twins of cities. However, current deep learning models suffer from a significant generalization gap. Unsupervised Domain Adaptation methods have recently emerged to tackle this issue. While domain-invariant feature learning using Maximum Mean Discrepancy has shown promise for images due to its simplicity, its application remains unexplored in outdoor mobile mapping point clouds. Moreover, previous methods do not consider the class information, which can lead to suboptimal adaptation performance. We propose a new approach—Contrastive Maximum Mean Discrepancy—to maximize intra-class domain alignment and minimize inter-class domain discrepancy, and integrate it into a 3D semantic segmentation model for LiDAR point clouds. The evaluation of our method with large-scale UDA datasets shows that it surpasses state-of-the-art UDA approaches for 3D LiDAR point clouds. CMMD is a promising UDA approach with strong potential for point cloud semantic segmentation.</div></div>","PeriodicalId":50633,"journal":{"name":"Computer Vision and Image Understanding","volume":"259 ","pages":"Article 104438"},"PeriodicalIF":4.3000,"publicationDate":"2025-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computer Vision and Image Understanding","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1077314225001614","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

Abstract

Semantic segmentation of 3D LiDAR point clouds is very important for applications like autonomous driving and digital twins of cities. However, current deep learning models suffer from a significant generalization gap. Unsupervised Domain Adaptation methods have recently emerged to tackle this issue. While domain-invariant feature learning using Maximum Mean Discrepancy has shown promise for images due to its simplicity, its application remains unexplored in outdoor mobile mapping point clouds. Moreover, previous methods do not consider the class information, which can lead to suboptimal adaptation performance. We propose a new approach—Contrastive Maximum Mean Discrepancy—to maximize intra-class domain alignment and minimize inter-class domain discrepancy, and integrate it into a 3D semantic segmentation model for LiDAR point clouds. The evaluation of our method with large-scale UDA datasets shows that it surpasses state-of-the-art UDA approaches for 3D LiDAR point clouds. CMMD is a promising UDA approach with strong potential for point cloud semantic segmentation.

Abstract Image

三维激光雷达分割中区域自适应的分布感知对比学习
三维激光雷达点云的语义分割对于自动驾驶和城市数字孪生等应用非常重要。然而,目前的深度学习模型存在显著的泛化差距。最近出现了无监督域自适应方法来解决这个问题。虽然使用最大平均差异的域不变特征学习由于其简单性而显示出对图像的希望,但其在室外移动地图点云中的应用仍未被探索。此外,以前的方法没有考虑类信息,这可能导致自适应性能不理想。我们提出了一种新的方法——对比最大平均差异——以最大化类内区域对齐和最小化类间区域差异,并将其集成到激光雷达点云的三维语义分割模型中。对我们的方法进行大规模UDA数据集的评估表明,它超过了3D激光雷达点云的最先进的UDA方法。CMMD是一种很有前途的UDA方法,具有很强的点云语义分割潜力。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
Computer Vision and Image Understanding
Computer Vision and Image Understanding 工程技术-工程:电子与电气
CiteScore
7.80
自引率
4.40%
发文量
112
审稿时长
79 days
期刊介绍: The central focus of this journal is the computer analysis of pictorial information. Computer Vision and Image Understanding publishes papers covering all aspects of image analysis from the low-level, iconic processes of early vision to the high-level, symbolic processes of recognition and interpretation. A wide range of topics in the image understanding area is covered, including papers offering insights that differ from predominant views. Research Areas Include: • Theory • Early vision • Data structures and representations • Shape • Range • Motion • Matching and recognition • Architecture and languages • Vision systems
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信