Alligat0R: Pre-Training through Covisibility Segmentation for Relative Camera Pose Regression | Read Paper on Bytez