{"title":"COLA: COarse-LAbel Multisource LiDAR Semantic Segmentation for Autonomous Driving","authors":"Jules Sanchez;Jean-Emmanuel Deschaud;François Goulette","doi":"10.1109/TRO.2025.3543302","DOIUrl":null,"url":null,"abstract":"LiDAR semantic segmentation (LSS) for autonomous driving has been a growing field of interest in recent years. Datasets and methods have appeared and expanded very quickly, but methods have not been updated to exploit this new data availability and rely on the same classical datasets. Different ways of performing LSS training and inference can be divided into several subfields, which include the following: domain generalization, source-to-source segmentation, and pretraining. In this work, we aim to improve results in all of these subfields with the novel approach of multisource training. Multisource training relies on the availability of various datasets at training time. To overcome the common obstacles in multisource training, we introduce the coarse labels and call the newly created multisource dataset COLA. We propose three applications of this new dataset that display systematic improvement over single-source strategies: COLA-DG for domain generalization (+10% ), COLA-S2S for source-to-source segmentation (+5.3% ), and COLA-PT for pretraining (+12% ). We demonstrate that multisource approaches bring systematic improvement over single-source approaches.","PeriodicalId":50388,"journal":{"name":"IEEE Transactions on Robotics","volume":"41 ","pages":"1742-1754"},"PeriodicalIF":9.4000,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Robotics","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/10891936/","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ROBOTICS","Score":null,"Total":0}
引用次数: 0
Abstract
LiDAR semantic segmentation (LSS) for autonomous driving has been a growing field of interest in recent years. Datasets and methods have appeared and expanded very quickly, but methods have not been updated to exploit this new data availability and rely on the same classical datasets. Different ways of performing LSS training and inference can be divided into several subfields, which include the following: domain generalization, source-to-source segmentation, and pretraining. In this work, we aim to improve results in all of these subfields with the novel approach of multisource training. Multisource training relies on the availability of various datasets at training time. To overcome the common obstacles in multisource training, we introduce the coarse labels and call the newly created multisource dataset COLA. We propose three applications of this new dataset that display systematic improvement over single-source strategies: COLA-DG for domain generalization (+10% ), COLA-S2S for source-to-source segmentation (+5.3% ), and COLA-PT for pretraining (+12% ). We demonstrate that multisource approaches bring systematic improvement over single-source approaches.
期刊介绍:
The IEEE Transactions on Robotics (T-RO) is dedicated to publishing fundamental papers covering all facets of robotics, drawing on interdisciplinary approaches from computer science, control systems, electrical engineering, mathematics, mechanical engineering, and beyond. From industrial applications to service and personal assistants, surgical operations to space, underwater, and remote exploration, robots and intelligent machines play pivotal roles across various domains, including entertainment, safety, search and rescue, military applications, agriculture, and intelligent vehicles.
Special emphasis is placed on intelligent machines and systems designed for unstructured environments, where a significant portion of the environment remains unknown and beyond direct sensing or control.