SLPOD: superclass learning on point cloud object detection

Abstract In the realm of point cloud object detection, classification tasks emphasize extracting common features to enhance generalization, often at the expense of individual-specific features. This limitation becomes particularly evident when handling intricate datasets like KITTI. Traditional mode...

Full description

Saved in:
Bibliographic Details
Main Authors: Xiaokang Yang, Kai Zhang, Yangyue Feng, Beibei Su, Yiming Cai, Kaibo Zhang, Zhiheng Zhang
Format: Article
Language:English
Published: Springer 2025-03-01
Series:Complex & Intelligent Systems
Subjects:
Online Access:https://doi.org/10.1007/s40747-025-01781-4
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850042429485350912
author Xiaokang Yang
Kai Zhang
Yangyue Feng
Beibei Su
Yiming Cai
Kaibo Zhang
Zhiheng Zhang
author_facet Xiaokang Yang
Kai Zhang
Yangyue Feng
Beibei Su
Yiming Cai
Kaibo Zhang
Zhiheng Zhang
author_sort Xiaokang Yang
collection DOAJ
description Abstract In the realm of point cloud object detection, classification tasks emphasize extracting common features to enhance generalization, often at the expense of individual-specific features. This limitation becomes particularly evident when handling intricate datasets like KITTI. Traditional models struggle to adequately capture individual-specific features, resulting in a scattered distribution of samples within the feature space and compromising the precision of object bounding boxes. To tackle this challenge, we introduce SLPOD, a Superclass-based point cloud object detection algorithm. Employing a siamese network structure, SLPOD conducts unsupervised clustering of samples within the same category to enhance the extraction of individual-specific features, thereby improving detection accuracy when confronted with complex datasets. Additionally, our approach integrates strategies such as voxel and point cloud feature fusion, global feature acquisition, and dynamic adjustment of sampling rates based on point sparsity, further enhancing the network’s capability to extract features. Experimental results demonstrate that SLPOD outperforms baseline algorithms in mean Average Precision on both KITTI and Waymo datasets, exhibiting robustness across diverse scenarios.
format Article
id doaj-art-a815f6bc06064cf7b6e2367ea19374f2
institution DOAJ
issn 2199-4536
2198-6053
language English
publishDate 2025-03-01
publisher Springer
record_format Article
series Complex & Intelligent Systems
spelling doaj-art-a815f6bc06064cf7b6e2367ea19374f22025-08-20T02:55:32ZengSpringerComplex & Intelligent Systems2199-45362198-60532025-03-0111511510.1007/s40747-025-01781-4SLPOD: superclass learning on point cloud object detectionXiaokang Yang0Kai Zhang1Yangyue Feng2Beibei Su3Yiming Cai4Kaibo Zhang5Zhiheng Zhang6School of Information Engineering, Changchun University of Finance and EconomicsSchool of Information Engineering, Changchun University of Finance and EconomicsSchool of Computer Science and Engineering, Changchun University of TechnologySchool of Computer Science and Engineering, Changchun University of TechnologySchool of Information Engineering, Changchun University of Finance and EconomicsSchool of Information Engineering, Changchun University of Finance and EconomicsSchool of Information Engineering, Changchun University of Finance and EconomicsAbstract In the realm of point cloud object detection, classification tasks emphasize extracting common features to enhance generalization, often at the expense of individual-specific features. This limitation becomes particularly evident when handling intricate datasets like KITTI. Traditional models struggle to adequately capture individual-specific features, resulting in a scattered distribution of samples within the feature space and compromising the precision of object bounding boxes. To tackle this challenge, we introduce SLPOD, a Superclass-based point cloud object detection algorithm. Employing a siamese network structure, SLPOD conducts unsupervised clustering of samples within the same category to enhance the extraction of individual-specific features, thereby improving detection accuracy when confronted with complex datasets. Additionally, our approach integrates strategies such as voxel and point cloud feature fusion, global feature acquisition, and dynamic adjustment of sampling rates based on point sparsity, further enhancing the network’s capability to extract features. Experimental results demonstrate that SLPOD outperforms baseline algorithms in mean Average Precision on both KITTI and Waymo datasets, exhibiting robustness across diverse scenarios.https://doi.org/10.1007/s40747-025-01781-4Point cloud object detectionContrastive learningUnsupervised learningDeep learning
spellingShingle Xiaokang Yang
Kai Zhang
Yangyue Feng
Beibei Su
Yiming Cai
Kaibo Zhang
Zhiheng Zhang
SLPOD: superclass learning on point cloud object detection
Complex & Intelligent Systems
Point cloud object detection
Contrastive learning
Unsupervised learning
Deep learning
title SLPOD: superclass learning on point cloud object detection
title_full SLPOD: superclass learning on point cloud object detection
title_fullStr SLPOD: superclass learning on point cloud object detection
title_full_unstemmed SLPOD: superclass learning on point cloud object detection
title_short SLPOD: superclass learning on point cloud object detection
title_sort slpod superclass learning on point cloud object detection
topic Point cloud object detection
Contrastive learning
Unsupervised learning
Deep learning
url https://doi.org/10.1007/s40747-025-01781-4
work_keys_str_mv AT xiaokangyang slpodsuperclasslearningonpointcloudobjectdetection
AT kaizhang slpodsuperclasslearningonpointcloudobjectdetection
AT yangyuefeng slpodsuperclasslearningonpointcloudobjectdetection
AT beibeisu slpodsuperclasslearningonpointcloudobjectdetection
AT yimingcai slpodsuperclasslearningonpointcloudobjectdetection
AT kaibozhang slpodsuperclasslearningonpointcloudobjectdetection
AT zhihengzhang slpodsuperclasslearningonpointcloudobjectdetection