Scene-aware Learning Network for Radar Object Detection

TL;DR

This paper introduces a scene-aware radar learning framework for robust object detection in autonomous driving, utilizing scene-specific branches, multiple architectures, and novel augmentation techniques to improve detection accuracy under various conditions.

Contribution

It proposes a novel scene-aware radar detection framework with multiple architectures, ensemble learning, and scene-specific augmentation for enhanced robustness and accuracy.

Findings

Achieved 75.0% average precision and 81.0% recall in ROD2021 Challenge.

Ranked first in parking lot scene with 97.8% precision and 98.6% recall.

Demonstrated robustness and effectiveness in diverse driving scenarios.

Abstract

Object detection is essential to safe autonomous or assisted driving. Previous works usually utilize RGB images or LiDAR point clouds to identify and localize multiple objects in self-driving. However, cameras tend to fail in bad driving conditions, e.g. bad weather or weak lighting, while LiDAR scanners are too expensive to get widely deployed in commercial applications. Radar has been drawing more and more attention due to its robustness and low cost. In this paper, we propose a scene-aware radar learning framework for accurate and robust object detection. First, the learning framework contains branches conditioning on the scene category of the radar sequence; with each branch optimized for a specific type of scene. Second, three different 3D autoencoder-based architectures are proposed for radar object detection and ensemble learning is performed over the different architectures to further boost the final performance. Third, we propose novel scene-aware sequence mix augmentation (SceneMix) and scene-specific post-processing to generate more robust detection results. In the ROD2021 Challenge, we achieved a final result of average precision of 75.0% and an average recall of 81.0%. Moreover, in the parking lot scene, our framework ranks first with an average precision of 97.8% and an average recall of 98.6%, which demonstrates the effectiveness of our framework.