A weakly supervised method for 3D object detection with partially annotated samples

In numerous practical applications, particularly in the field of autonomous driving, acquiring annotated datasets that include both images and LiDAR point clouds simultaneously presents significant challenges and incurs substantial costs. To overcome the limitations of limited sample annotations, we...

Full description

Saved in:
Bibliographic Details
Main Authors: Bin Lu, Qing Li, Yanju Liang
Format: Article
Language:English
Published: SAGE Publishing 2025-04-01
Series:Measurement + Control
Online Access:https://doi.org/10.1177/00202940241297568
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850093910222700544
author Bin Lu
Qing Li
Yanju Liang
author_facet Bin Lu
Qing Li
Yanju Liang
author_sort Bin Lu
collection DOAJ
description In numerous practical applications, particularly in the field of autonomous driving, acquiring annotated datasets that include both images and LiDAR point clouds simultaneously presents significant challenges and incurs substantial costs. To overcome the limitations of limited sample annotations, we propose an innovative weakly supervised learning methodology that utilizes reciprocal knowledge transfer between image detection models and 3D point cloud detection models. To the best of our knowledge, this area has not been explored by prior research teams. Our approach effectively addresses the alignment challenge of diverse modal features from an aerial perspective. Through heatmap prediction, we successfully facilitate knowledge transfer between the image detection and 3D point cloud detection models. Additionally, we conduct extensive experiments to evaluate the performance of our models under different parameters in the domain adaptation process, employing Exponential Moving Average (EMA) progressive learning. Furthermore, we explore the advantages of incorporating regression and prediction fusion heads to enhance weakly supervised learning. Remarkably, our experimental results on the widely accessible KITTI datasets demonstrate that our proposed approach achieves outstanding performance in 3D object detection under weak supervision, surpassing the baseline performance of the original 3D point cloud detection model.
format Article
id doaj-art-3fdee36b159448b9ab27e70f4adb74c8
institution DOAJ
issn 0020-2940
language English
publishDate 2025-04-01
publisher SAGE Publishing
record_format Article
series Measurement + Control
spelling doaj-art-3fdee36b159448b9ab27e70f4adb74c82025-08-20T02:41:48ZengSAGE PublishingMeasurement + Control0020-29402025-04-015810.1177/00202940241297568A weakly supervised method for 3D object detection with partially annotated samplesBin Lu0Qing Li1Yanju Liang2Wuxi IoT Innovation Center Co., Wuxi, ChinaInstitute of Microelectronics of the Chinese Academy of Sciences,Beijing 100029, ChinaWuxi IoT Innovation Center Co., Wuxi, ChinaIn numerous practical applications, particularly in the field of autonomous driving, acquiring annotated datasets that include both images and LiDAR point clouds simultaneously presents significant challenges and incurs substantial costs. To overcome the limitations of limited sample annotations, we propose an innovative weakly supervised learning methodology that utilizes reciprocal knowledge transfer between image detection models and 3D point cloud detection models. To the best of our knowledge, this area has not been explored by prior research teams. Our approach effectively addresses the alignment challenge of diverse modal features from an aerial perspective. Through heatmap prediction, we successfully facilitate knowledge transfer between the image detection and 3D point cloud detection models. Additionally, we conduct extensive experiments to evaluate the performance of our models under different parameters in the domain adaptation process, employing Exponential Moving Average (EMA) progressive learning. Furthermore, we explore the advantages of incorporating regression and prediction fusion heads to enhance weakly supervised learning. Remarkably, our experimental results on the widely accessible KITTI datasets demonstrate that our proposed approach achieves outstanding performance in 3D object detection under weak supervision, surpassing the baseline performance of the original 3D point cloud detection model.https://doi.org/10.1177/00202940241297568
spellingShingle Bin Lu
Qing Li
Yanju Liang
A weakly supervised method for 3D object detection with partially annotated samples
Measurement + Control
title A weakly supervised method for 3D object detection with partially annotated samples
title_full A weakly supervised method for 3D object detection with partially annotated samples
title_fullStr A weakly supervised method for 3D object detection with partially annotated samples
title_full_unstemmed A weakly supervised method for 3D object detection with partially annotated samples
title_short A weakly supervised method for 3D object detection with partially annotated samples
title_sort weakly supervised method for 3d object detection with partially annotated samples
url https://doi.org/10.1177/00202940241297568
work_keys_str_mv AT binlu aweaklysupervisedmethodfor3dobjectdetectionwithpartiallyannotatedsamples
AT qingli aweaklysupervisedmethodfor3dobjectdetectionwithpartiallyannotatedsamples
AT yanjuliang aweaklysupervisedmethodfor3dobjectdetectionwithpartiallyannotatedsamples
AT binlu weaklysupervisedmethodfor3dobjectdetectionwithpartiallyannotatedsamples
AT qingli weaklysupervisedmethodfor3dobjectdetectionwithpartiallyannotatedsamples
AT yanjuliang weaklysupervisedmethodfor3dobjectdetectionwithpartiallyannotatedsamples