SparseRadNet: Sparse Perception Neural Network on Subsampled Radar Data

TL;DR

SparseRadNet introduces an adaptive subsampling and a specialized neural network architecture to effectively process sparse radar data, improving perception accuracy in autonomous driving tasks while reducing data requirements.

Contribution

The paper presents a novel subsampling method and a dual-branch neural network architecture designed specifically for sparse radar data in perception tasks.

Findings

Outperforms state-of-the-art in object detection on RADIal dataset.

Achieves near state-of-the-art in freespace segmentation with sparse data.

Demonstrates effective exploitation of radar data sparsity patterns.

Abstract

Radar-based perception has gained increasing attention in autonomous driving, yet the inherent sparsity of radars poses challenges. Radar raw data often contains excessive noise, whereas radar point clouds retain only limited information. In this work, we holistically treat the sparse nature of radar data by introducing an adaptive subsampling method together with a tailored network architecture that exploits the sparsity patterns to discover global and local dependencies in the radar signal. Our subsampling module selects a subset of pixels from range-doppler (RD) spectra that contribute most to the downstream perception tasks. To improve the feature extraction on sparse subsampled data, we propose a new way of applying graph neural networks on radar data and design a novel two-branch backbone to capture both global and local neighbor information. An attentive fusion module is applied to combine features from both branches. Experiments on the RADIal dataset show that our SparseRadNet exceeds state-of-the-art (SOTA) performance in object detection and achieves close to SOTA accuracy in freespace segmentation, meanwhile using sparse subsampled input data.