RoRE: Rotary Ray Embedding for Generalised Multi-Modal Scene Understanding → February 6, 2026 Categories Publication Previous RectConv: Adapting CNNs for Fisheye Cameras without Retraining
