SegLoc: Learning Segmentation-Based Representations for Privacy-Preserving Visual Localization | Read Paper on Bytez