TransRadar: Adaptive-Directional Transformer for Real-Time Multi-View Radar Semantic Segmentation
This work addresses scene understanding for autonomous driving by improving radar perception, which is low-cost and weather-resistant, but it is incremental as it builds on existing radar segmentation methods.
The paper tackles radar-based semantic segmentation for autonomous driving by proposing TransRadar, a novel architecture with adaptive attention and tailored loss functions, which outperforms state-of-the-art methods on CARRADA and RADIal datasets with smaller model sizes.
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