TransRadar: Adaptive-Directional Transformer for Real-Time Multi-View Radar Semantic Segmentation

TL;DR

TransRadar introduces an adaptive-directional transformer architecture for real-time multi-view radar semantic segmentation, effectively handling radar data noise and sparsity, and outperforming existing methods on key datasets.

Contribution

It presents a novel multi-input fusion architecture with specialized attention blocks and loss functions tailored for radar data, advancing radar scene understanding.

Findings

Outperforms state-of-the-art on CARRADA and RADIal datasets

Achieves higher accuracy with smaller model size

Handles radar data noise and sparsity effectively

Abstract

Scene understanding plays an essential role in enabling autonomous driving and maintaining high standards of performance and safety. To address this task, cameras and laser scanners (LiDARs) have been the most commonly used sensors, with radars being less popular. Despite that, radars remain low-cost, information-dense, and fast-sensing techniques that are resistant to adverse weather conditions. While multiple works have been previously presented for radar-based scene semantic segmentation, the nature of the radar data still poses a challenge due to the inherent noise and sparsity, as well as the disproportionate foreground and background. In this work, we propose a novel approach to the semantic segmentation of radar scenes using a multi-input fusion of radar data through a novel architecture and loss functions that are tailored to tackle the drawbacks of radar perception. Our novel architecture includes an efficient attention block that adaptively captures important feature information. Our method, TransRadar, outperforms state-of-the-art methods on the CARRADA and RADIal datasets while having smaller model sizes. https://github.com/YahiDar/TransRadar