MSDNet: Efficient 4D Radar Super-Resolution via Multi-Stage Distillation

2025-09-16

AI Summary

4D radar point clouds suffer from sparsity, high noise levels, and geometric inconsistency, making it challenging to achieve both high accuracy and efficiency in super-resolution reconstruction. Method: This paper proposes a multi-stage feature distillation framework featuring: (i) a two-stage distillation mechanism that jointly leverages reconstruction-guided and diffusion-guided supervision; (ii) a noise adapter that dynamically aligns radar feature representations with their inherent noise characteristics; and (iii) a lightweight diffusion network coupled with feature-reconstruction alignment to enable effective radar–LiDAR cross-modal knowledge transfer. Results: Evaluated on the VoD dataset and a custom radar dataset, our method significantly improves reconstruction fidelity and downstream task performance (e.g., 3D object detection), reduces inference latency by 42%, and demonstrates superior generalization over existing diffusion-based approaches—achieving a more favorable accuracy–efficiency trade-off.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Feature Construction/Reformulation
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

4D radar super-resolution, which aims to reconstruct sparse and noisy point clouds into dense and geometrically consistent representations, is a foundational problem in autonomous perception. However, existing methods often suffer from high training cost or rely on complex diffusion-based sampling, resulting in high inference latency and poor generalization, making it difficult to balance accuracy and efficiency. To address these limitations, we propose MSDNet, a multi-stage distillation framework that efficiently transfers dense LiDAR priors to 4D radar features to achieve both high reconstruction quality and computational efficiency. The first stage performs reconstruction-guided feature distillation, aligning and densifying the student's features through feature reconstruction. In the second stage, we propose diffusion-guided feature distillation, which treats the stage-one distilled features as a noisy version of the teacher's representations and refines them via a lightweight diffusion network. Furthermore, we introduce a noise adapter that adaptively aligns the noise level of the feature with a predefined diffusion timestep, enabling a more precise denoising. Extensive experiments on the VoD and in-house datasets demonstrate that MSDNet achieves both high-fidelity reconstruction and low-latency inference in the task of 4D radar point cloud super-resolution, and consistently improves performance on downstream tasks. The code will be publicly available upon publication.

Problem

• Reconstruct sparse noisy 4D radar point clouds
• Balance accuracy efficiency super-resolution training
• Reduce inference latency improve generalization performance

Innovation

• Multi-stage distillation for 4D radar super-resolution
• Reconstruction-guided feature distillation alignment
• Diffusion-guided refinement with adaptive noise adapter