DSFEC: Efficient and Deployable Deep Radar Object Detection

TL;DR

This paper introduces DSFEC, a radar object detection model optimized for edge devices, utilizing depthwise separable convolutions and a novel feature module to enhance performance and deployability.

Contribution

The paper proposes DSFEC with a new feature enhancement module, achieving significant efficiency improvements and better performance on resource-limited edge hardware.

Findings

DSFEC-M improves performance by 14.6% with 60% fewer GFLOPs.

DSFEC-S reduces GFLOPs by 78.5% with a 3.76% performance gain.

Runtime on Raspberry Pi is reduced by 74.5%.

Abstract

Deploying radar object detection models on resource-constrained edge devices like the Raspberry Pi poses significant challenges due to the large size of the model and the limited computational power and the memory of the Pi. In this work, we explore the efficiency of Depthwise Separable Convolutions in radar object detection networks and integrate them into our model. Additionally, we introduce a novel Feature Enhancement and Compression (FEC) module to the PointPillars feature encoder to further improve the model performance. With these innovations, we propose the DSFEC-L model and its two versions, which outperform the baseline (23.9 mAP of Car class, 20.72 GFLOPs) on nuScenes dataset: 1). An efficient DSFEC-M model with a 14.6% performance improvement and a 60% reduction in GFLOPs. 2). A deployable DSFEC-S model with a 3.76% performance improvement and a remarkable 78.5% reduction in GFLOPs. Despite marginal performance gains, our deployable model achieves an impressive 74.5% reduction in runtime on the Raspberry Pi compared to the baseline.