Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework

We investigate the problem of 3D point clouds semantic segmentation. Recently, a large amount of research work has focused on local feature aggregation. However, the foundational framework of semantic segmentation of 3D point clouds has been neglected, where the majority of current methods default t...

Full description

Saved in:
Bibliographic Details
Main Authors: Ziyin Zeng, Qingyong Hu, Zhong Xie, Bijun Li, Jian Zhou, Yongyang Xu
Format: Article
Language:English
Published: Elsevier 2025-02-01
Series:International Journal of Applied Earth Observations and Geoinformation
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1569843224006678
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849720344673255424
author Ziyin Zeng
Qingyong Hu
Zhong Xie
Bijun Li
Jian Zhou
Yongyang Xu
author_facet Ziyin Zeng
Qingyong Hu
Zhong Xie
Bijun Li
Jian Zhou
Yongyang Xu
author_sort Ziyin Zeng
collection DOAJ
description We investigate the problem of 3D point clouds semantic segmentation. Recently, a large amount of research work has focused on local feature aggregation. However, the foundational framework of semantic segmentation of 3D point clouds has been neglected, where the majority of current methods default to the U-Net framework. In this study, we present U-Next, a small but mighty framework designed specifically for point cloud semantic segmentation. The key innovation of this framework is to capture multi-scale hierarchical features. Specifically, we construct the U-Next by stacking multiple U-Net L1 sub-networks in a dense arrangement to diminish the semantic gap. Concurrently, it integrates feature maps across various scales to proficiently restore intricate fine-grained details. Additionally, a multi-level deep supervision mechanism is introduced for smoothing gradient propagation and facilitating network optimization. We conduct extensive experiments on benchmarks, including the indoor S3DIS dataset, the LiDAR-based outdoor Toronto3D dataset, and the urban-scale photogrammetry-based SensatUrban dataset, demonstrate the superiority of U-Next. The U-Next framework consistently exhibits significant performance enhancements across various benchmarks and baselines, demonstrating its considerable potential as a versatile point-based framework for future endeavors. The code has been released at https://github.com/zeng-ziyin/U-Next.
format Article
id doaj-art-b13bdfa89b964da8b611d1ca42271932
institution DOAJ
issn 1569-8432
language English
publishDate 2025-02-01
publisher Elsevier
record_format Article
series International Journal of Applied Earth Observations and Geoinformation
spelling doaj-art-b13bdfa89b964da8b611d1ca422719322025-08-20T03:11:57ZengElsevierInternational Journal of Applied Earth Observations and Geoinformation1569-84322025-02-0113610430910.1016/j.jag.2024.104309Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next frameworkZiyin Zeng0Qingyong Hu1Zhong Xie2Bijun Li3Jian Zhou4Yongyang Xu5School of Computer Science, China University of Geosciences, Wuhan, China; State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, ChinaAcademy of Military Sciences, Beijing, ChinaSchool of Computer Science, China University of Geosciences, Wuhan, ChinaState Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, ChinaState Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, ChinaSchool of Computer Science, China University of Geosciences, Wuhan, China; Corresponding author.We investigate the problem of 3D point clouds semantic segmentation. Recently, a large amount of research work has focused on local feature aggregation. However, the foundational framework of semantic segmentation of 3D point clouds has been neglected, where the majority of current methods default to the U-Net framework. In this study, we present U-Next, a small but mighty framework designed specifically for point cloud semantic segmentation. The key innovation of this framework is to capture multi-scale hierarchical features. Specifically, we construct the U-Next by stacking multiple U-Net L1 sub-networks in a dense arrangement to diminish the semantic gap. Concurrently, it integrates feature maps across various scales to proficiently restore intricate fine-grained details. Additionally, a multi-level deep supervision mechanism is introduced for smoothing gradient propagation and facilitating network optimization. We conduct extensive experiments on benchmarks, including the indoor S3DIS dataset, the LiDAR-based outdoor Toronto3D dataset, and the urban-scale photogrammetry-based SensatUrban dataset, demonstrate the superiority of U-Next. The U-Next framework consistently exhibits significant performance enhancements across various benchmarks and baselines, demonstrating its considerable potential as a versatile point-based framework for future endeavors. The code has been released at https://github.com/zeng-ziyin/U-Next.http://www.sciencedirect.com/science/article/pii/S1569843224006678Point cloudSemantic segmentationNetwork architectureDeep supervision
spellingShingle Ziyin Zeng
Qingyong Hu
Zhong Xie
Bijun Li
Jian Zhou
Yongyang Xu
Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework
International Journal of Applied Earth Observations and Geoinformation
Point cloud
Semantic segmentation
Network architecture
Deep supervision
title Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework
title_full Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework
title_fullStr Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework
title_full_unstemmed Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework
title_short Small but mighty: Enhancing 3D point clouds semantic segmentation with U-Next framework
title_sort small but mighty enhancing 3d point clouds semantic segmentation with u next framework
topic Point cloud
Semantic segmentation
Network architecture
Deep supervision
url http://www.sciencedirect.com/science/article/pii/S1569843224006678
work_keys_str_mv AT ziyinzeng smallbutmightyenhancing3dpointcloudssemanticsegmentationwithunextframework
AT qingyonghu smallbutmightyenhancing3dpointcloudssemanticsegmentationwithunextframework
AT zhongxie smallbutmightyenhancing3dpointcloudssemanticsegmentationwithunextframework
AT bijunli smallbutmightyenhancing3dpointcloudssemanticsegmentationwithunextframework
AT jianzhou smallbutmightyenhancing3dpointcloudssemanticsegmentationwithunextframework
AT yongyangxu smallbutmightyenhancing3dpointcloudssemanticsegmentationwithunextframework