Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths

This paper introduces a depth-guided 3D object detection method that enhances the feature extraction capability of the backbone network through weak supervision. It combines large kernel convolution, global response normalization, and layer normalization techniques to significantly improve feature r...

Full description

Saved in:
Bibliographic Details
Main Authors: Xinwang Zheng, Guangsong Yang, Lu Yang, Chengyu Lu
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10740295/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846170462202101760
author Xinwang Zheng
Guangsong Yang
Lu Yang
Chengyu Lu
author_facet Xinwang Zheng
Guangsong Yang
Lu Yang
Chengyu Lu
author_sort Xinwang Zheng
collection DOAJ
description This paper introduces a depth-guided 3D object detection method that enhances the feature extraction capability of the backbone network through weak supervision. It combines large kernel convolution, global response normalization, and layer normalization techniques to significantly improve feature robustness under weakly supervised conditions. Additionally, the depth estimation module’s feature extraction ability is bolstered by optimizing the depth-guided encoder and incorporating large-kernel depthwise separable convolutions alongside a spatial attention mechanism. On the decoder side, deformable convolutions are employed to modulate deep feature maps, reducing inference and training time while minimizing model complexity. This approach avoids the complexity associated with transformer architectures. Experiments on the KITTI 3D dataset demonstrate that the method diminishes reliance on manual labeling and can notably enhance detection accuracy while simultaneously improving processing speed.
format Article
id doaj-art-d1e58b1c32984e2aa573e007519fb9c6
institution Kabale University
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-d1e58b1c32984e2aa573e007519fb9c62024-11-12T00:01:27ZengIEEEIEEE Access2169-35362024-01-011216299016300010.1109/ACCESS.2024.348874810740295Constructing 3D Object Detectors Based on Deformable Convolutional Guided DepthsXinwang Zheng0Guangsong Yang1https://orcid.org/0000-0002-1489-9841Lu Yang2https://orcid.org/0000-0002-5395-8912Chengyu Lu3https://orcid.org/0009-0003-6180-8441Chengyi College, Jimei University, Xiamen, ChinaSchool of Ocean Information Engineering, Jimei University, Xiamen, ChinaChengyi College, Jimei University, Xiamen, ChinaChengyi College, Jimei University, Xiamen, ChinaThis paper introduces a depth-guided 3D object detection method that enhances the feature extraction capability of the backbone network through weak supervision. It combines large kernel convolution, global response normalization, and layer normalization techniques to significantly improve feature robustness under weakly supervised conditions. Additionally, the depth estimation module’s feature extraction ability is bolstered by optimizing the depth-guided encoder and incorporating large-kernel depthwise separable convolutions alongside a spatial attention mechanism. On the decoder side, deformable convolutions are employed to modulate deep feature maps, reducing inference and training time while minimizing model complexity. This approach avoids the complexity associated with transformer architectures. Experiments on the KITTI 3D dataset demonstrate that the method diminishes reliance on manual labeling and can notably enhance detection accuracy while simultaneously improving processing speed.https://ieeexplore.ieee.org/document/10740295/3D object detectiondeep learningdepth guidanceautonomous automobilesmodel refinement
spellingShingle Xinwang Zheng
Guangsong Yang
Lu Yang
Chengyu Lu
Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths
IEEE Access
3D object detection
deep learning
depth guidance
autonomous automobiles
model refinement
title Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths
title_full Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths
title_fullStr Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths
title_full_unstemmed Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths
title_short Constructing 3D Object Detectors Based on Deformable Convolutional Guided Depths
title_sort constructing 3d object detectors based on deformable convolutional guided depths
topic 3D object detection
deep learning
depth guidance
autonomous automobiles
model refinement
url https://ieeexplore.ieee.org/document/10740295/
work_keys_str_mv AT xinwangzheng constructing3dobjectdetectorsbasedondeformableconvolutionalguideddepths
AT guangsongyang constructing3dobjectdetectorsbasedondeformableconvolutionalguideddepths
AT luyang constructing3dobjectdetectorsbasedondeformableconvolutionalguideddepths
AT chengyulu constructing3dobjectdetectorsbasedondeformableconvolutionalguideddepths