Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

V2X-DGPE: Addressing Domain Gaps and Pose Errors for Robust Collaborative 3D Object Detection

Sichao Wang, Ming Yuan, Chuang Zhang, Qing Xu, Lei He, Jianqiang Wang

TL;DR

V2X-DGPE tackles domain gaps and pose errors in cooperative 3D object detection by fusing vehicle and infrastructure data at the feature level. It combines a teacher–student knowledge distillation framework, a temporal fusion of current and historical BEV features, a lightweight feature compensation module, and a collaborative fusion module with heterogeneous and deformable self-attention to model cross-agent interactions while dynamically addressing misalignment. The approach yields state-of-the-art accuracy and robust performance under pose perturbations on the real-world DAIR-V2X dataset, outperforming prior intermediate fusion methods. By leveraging historical context and adaptive sampling, V2X-DGPE provides resilient, cross-domain perception suitable for practical V2X deployments.

Abstract

In V2X collaborative perception, the domain gaps between heterogeneous nodes pose a significant challenge for effective information fusion. Pose errors arising from latency and GPS localization noise further exacerbate the issue by leading to feature misalignment. To overcome these challenges, we propose V2X-DGPE, a high-accuracy and robust V2X feature-level collaborative perception framework. V2X-DGPE employs a Knowledge Distillation Framework and a Feature Compensation Module to learn domain-invariant representations from multi-source data, effectively reducing the feature distribution gap between vehicles and roadside infrastructure. Historical information is utilized to provide the model with a more comprehensive understanding of the current scene. Furthermore, a Collaborative Fusion Module leverages a heterogeneous self-attention mechanism to extract and integrate heterogeneous representations from vehicles and infrastructure. To address pose errors, V2X-DGPE introduces a deformable attention mechanism, enabling the model to adaptively focus on critical parts of the input features by dynamically offsetting sampling points. Extensive experiments on the real-world DAIR-V2X dataset demonstrate that the proposed method outperforms existing approaches, achieving state-of-the-art detection performance. The code is available at https://github.com/wangsch10/V2X-DGPE.

V2X-DGPE: Addressing Domain Gaps and Pose Errors for Robust Collaborative 3D Object Detection

TL;DR

V2X-DGPE tackles domain gaps and pose errors in cooperative 3D object detection by fusing vehicle and infrastructure data at the feature level. It combines a teacher–student knowledge distillation framework, a temporal fusion of current and historical BEV features, a lightweight feature compensation module, and a collaborative fusion module with heterogeneous and deformable self-attention to model cross-agent interactions while dynamically addressing misalignment. The approach yields state-of-the-art accuracy and robust performance under pose perturbations on the real-world DAIR-V2X dataset, outperforming prior intermediate fusion methods. By leveraging historical context and adaptive sampling, V2X-DGPE provides resilient, cross-domain perception suitable for practical V2X deployments.

Abstract

In V2X collaborative perception, the domain gaps between heterogeneous nodes pose a significant challenge for effective information fusion. Pose errors arising from latency and GPS localization noise further exacerbate the issue by leading to feature misalignment. To overcome these challenges, we propose V2X-DGPE, a high-accuracy and robust V2X feature-level collaborative perception framework. V2X-DGPE employs a Knowledge Distillation Framework and a Feature Compensation Module to learn domain-invariant representations from multi-source data, effectively reducing the feature distribution gap between vehicles and roadside infrastructure. Historical information is utilized to provide the model with a more comprehensive understanding of the current scene. Furthermore, a Collaborative Fusion Module leverages a heterogeneous self-attention mechanism to extract and integrate heterogeneous representations from vehicles and infrastructure. To address pose errors, V2X-DGPE introduces a deformable attention mechanism, enabling the model to adaptively focus on critical parts of the input features by dynamically offsetting sampling points. Extensive experiments on the real-world DAIR-V2X dataset demonstrate that the proposed method outperforms existing approaches, achieving state-of-the-art detection performance. The code is available at https://github.com/wangsch10/V2X-DGPE.
Paper Structure (15 sections, 7 equations, 5 figures, 4 tables)

This paper contains 15 sections, 7 equations, 5 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. METHODOLOGY
  4. Overall Architecture
  5. Feature Extraction Module
  6. Temporal Fusion Module
  7. Feature Compensation Module
  8. Collaborative Fusion Module
  9. EXPERIMENTS
  10. DAIR-V2X
  11. Experimental Setup
  12. Performance Comparison
  13. Ablation Studies
  14. Pose Errors Reslut
  15. CONCLUSION

Figures (5)

  • Figure 1: A sample from Dair-V2X illustrating the domain gap between vehicle (40-line LiDAR) and infrastructure (300-line LiDAR) point clouds .
  • Figure 2: Overview architecture of V2X-DGPE. It employs a Knowledge Distillation Framework, comprising five key components arranged sequentially: BEV Feature Extraction Module, Temporal Fusion Module, Feature Compensation Module, Collaborative Fusion Module, and the Detection Head.
  • Figure 3: Illustration of the Temporal Fusion Module and Feature Compensation Module.
  • Figure 4: (a) The architecture of the Collaborative Fusion Module. (b) Illustration of the heterogeneous self-attention module. (c) Illustration of the deformable self-attention module.
  • Figure 5: Detection visualization of V2X-ViT, DI-V2X, Coalign, and V2X-DGPE under Gaussian noise with $\sigma_t = 0.4$m and $\sigma_r = 0.4^\circ$. The green boxes represent the ground truth, while the red boxes represent the detected results. Compared to other advanced models, the proposed model, V2X-DGPE, demonstrates superior detection accuracy, with its detection boxes being noticeably more precise.