This article discusses a new approach to simultaneously detect clutter and semantically segment moving objects in the data from automotive radar sensors. In contrast to the usual practice of handling these tasks separately with different neural network models, the authors propose using a single, jointly used model. This novel approach, which includes a new augmented multi-head architecture, is shown to be highly effective, outperforming existing networks for semantic segmentation on the RadarScenes dataset. The article highlights the key role of radar sensors in the perception system of autonomous vehicles, noting their robustness to adverse weather conditions.
Publication date: 14 Nov 2023
Project Page: N/A
Paper: https://arxiv.org/pdf/2311.07247
