AI RESEARCH
Weather-Robust Scene Semantics with Vision-Aligned 4D Radar
arXiv CS.CV
•
ArXi:2605.07367v1 Announce Type: cross Cameras and LiDAR degrade in rain, fog, and snow, while millimeter-wave radar remains largely unaffected. We align a radar encoder to frozen SigLIP vision embeddings and decode structured scene captions through a frozen vision-language model (VLM) with approximately 7M trainable parameters. On K-RADAR with held-out fog, light snow, and heavy snow sequences, all radar configurations outperform a camera baseline that collapses to over 90% hallucination.