DeepReflecs: Deep Learning for Automotive Object Classification with Radar Reflections

TL;DR

DeepReflecs introduces a lightweight deep learning approach for automotive radar data that accurately classifies objects like pedestrians, cyclists, and cars, bridging the gap between traditional handcrafted features and complex neural networks.

Contribution

It proposes a novel, efficient deep learning model tailored for radar reflection data, handling unordered inputs and combining local and global features for improved classification.

Findings

Outperforms existing handcrafted and learned feature methods

Handles unordered radar data effectively

Ablation study confirms the importance of the global context layer

Abstract

This paper presents an novel object type classification method for automotive applications which uses deep learning with radar reflections. The method provides object class information such as pedestrian, cyclist, car, or non-obstacle. The method is both powerful and efficient, by using a light-weight deep learning approach on reflection level radar data. It fills the gap between low-performant methods of handcrafted features and high-performant methods with convolutional neural networks. The proposed network exploits the specific characteristics of radar reflection data: It handles unordered lists of arbitrary length as input and it combines both extraction of local and global features. In experiments with real data the proposed network outperforms existing methods of handcrafted or learned features. An ablation study analyzes the impact of the proposed global context layer.