Graph Query Networks for Object Detection with Automotive Radar

TL;DR

This paper presents Graph Query Networks (GQN), an attention-based graph modeling framework for radar-based object detection that significantly improves accuracy and efficiency over prior methods.

Contribution

The paper introduces GQN with novel graph query mechanisms and modules, advancing radar object detection by capturing relational and contextual features more effectively.

Findings

GQN improves relative mAP by up to +53% on NuScenes.

GQN achieves +8.2% gain over previous radar methods.

Reduces peak graph construction overhead by 80%.

Abstract

Object detection with 3D radar is essential for 360-degree automotive perception, but radar's long wavelengths produce sparse and irregular reflections that challenge traditional grid and sequence-based convolutional and transformer detectors. This paper introduces Graph Query Networks (GQN), an attention-based framework that models objects sensed by radar as graphs, to extract individualized relational and contextual features. GQN employs a novel concept of graph queries to dynamically attend over the bird's-eye view (BEV) space, constructing object-specific graphs processed by two novel modules: EdgeFocus for relational reasoning and DeepContext Pooling for contextual aggregation. On the NuScenes dataset, GQN improves relative mAP by up to +53%, including a +8.2% gain over the strongest prior radar method, while reducing peak graph construction overhead by 80% with moderate FLOPs cost.