LDRFusion: A LiDAR-Dominant multimodal refinement framework for 3D object detection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Jijun, Wu, Yan, Mo, Yujian, Zhao, Junqiao, Yan, Jun, Hu, Yinghao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918131109199872
author Wang, Jijun
Wu, Yan
Mo, Yujian
Zhao, Junqiao
Yan, Jun
Hu, Yinghao
author_facet Wang, Jijun
Wu, Yan
Mo, Yujian
Zhao, Junqiao
Yan, Jun
Hu, Yinghao
contents Existing LiDAR-Camera fusion methods have achieved strong results in 3D object detection. To address the sparsity of point clouds, previous approaches typically construct spatial pseudo point clouds via depth completion as auxiliary input and adopts a proposal-refinement framework to generate detection results. However, introducing pseudo points inevitably brings noise, potentially resulting in inaccurate predictions. Considering the differing roles and reliability levels of each modality, we propose LDRFusion, a novel Lidar-dominant two-stage refinement framework for multi-sensor fusion. The first stage soley relies on LiDAR to produce accurately localized proposals, followed by a second stage where pseudo point clouds are incorporated to detect challenging instances. The instance-level results from both stages are subsequently merged. To further enhance the representation of local structures in pseudo point clouds, we present a hierarchical pseudo point residual encoding module, which encodes neighborhood sets using both feature and positional residuals. Experiments on the KITTI dataset demonstrate that our framework consistently achieves strong performance across multiple categories and difficulty levels.
format Preprint
id arxiv_https___arxiv_org_abs_2507_16224
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LDRFusion: A LiDAR-Dominant multimodal refinement framework for 3D object detection
Wang, Jijun
Wu, Yan
Mo, Yujian
Zhao, Junqiao
Yan, Jun
Hu, Yinghao
Computer Vision and Pattern Recognition
Existing LiDAR-Camera fusion methods have achieved strong results in 3D object detection. To address the sparsity of point clouds, previous approaches typically construct spatial pseudo point clouds via depth completion as auxiliary input and adopts a proposal-refinement framework to generate detection results. However, introducing pseudo points inevitably brings noise, potentially resulting in inaccurate predictions. Considering the differing roles and reliability levels of each modality, we propose LDRFusion, a novel Lidar-dominant two-stage refinement framework for multi-sensor fusion. The first stage soley relies on LiDAR to produce accurately localized proposals, followed by a second stage where pseudo point clouds are incorporated to detect challenging instances. The instance-level results from both stages are subsequently merged. To further enhance the representation of local structures in pseudo point clouds, we present a hierarchical pseudo point residual encoding module, which encodes neighborhood sets using both feature and positional residuals. Experiments on the KITTI dataset demonstrate that our framework consistently achieves strong performance across multiple categories and difficulty levels.
title LDRFusion: A LiDAR-Dominant multimodal refinement framework for 3D object detection
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2507.16224