4D-RaDiff: Latent Diffusion for 4D Radar Point Cloud Generation

TL;DR

This paper introduces 4D-RaDiff, a latent diffusion framework for generating realistic 4D radar point clouds to enhance training data for automotive perception, significantly reducing the need for annotated real radar data.

Contribution

The novel latent diffusion approach tailored for radar point clouds enables high-quality synthetic data generation, improving detection performance and reducing annotation requirements.

Findings

Synthetic radar data improves object detection accuracy.

Pre-training with generated data reduces annotation needs by up to 90%.

Method outperforms existing data augmentation techniques.

Abstract

Automotive radar has shown promising developments in environment perception due to its cost-effectiveness and robustness in adverse weather conditions. However, the limited availability of annotated radar data poses a significant challenge for advancing radar-based perception systems. To address this limitation, we propose a novel framework to generate 4D radar point clouds for training and evaluating object detectors. Unlike image-based diffusion, our method is designed to consider the sparsity and unique characteristics of radar point clouds by applying diffusion to a latent point cloud representation. Within this latent space, generation is controlled via conditioning at either the object or scene level. The proposed 4D-RaDiff converts unlabeled bounding boxes into high-quality radar annotations and transforms existing LiDAR point cloud data into realistic radar scenes. Experiments demonstrate that incorporating synthetic radar data of 4D-RaDiff as data augmentation method during training consistently improves object detection performance compared to training on real data only. In addition, pre-training on our synthetic data reduces the amount of required annotated radar data by up to 90% while achieving comparable object detection performance.