Tessellating The Earth

2026-06-25

AI Summary

Existing geocoding methods allocate representational capacity uniformly, struggling to meet the demands of high-information-density regions such as urban areas. This work proposes the first end-to-end learnable spherical Voronoi geocoder, which dynamically optimizes the distribution of representational resources through a differentiable mechanism and incorporates a global semantic token to integrate environmental semantic knowledge from remote sensing imagery. By combining learnable spherical Voronoi partitioning, differentiable embedding transfer, and semantic distillation, the method achieves state-of-the-art performance across multiple geospatial classification and regression tasks. Furthermore, it serves as a strong geographic prior, significantly improving fine-grained species classification accuracy on iNaturalist-2018.

Abstract

Geolocation encoders, which map geographic coordinates to learned representations, are emerging as an effective means of capturing visual and non-visual characteristics from a latitude-longitude pair alone. However, existing approaches project coordinates onto fixed bases (e.g., spherical harmonics), allocating representational capacity uniformly and devoting equal resources to the open ocean and to a developing city. We introduce Tessellating the Earth (TTE), a location encoder built from learnable Spherical Voronoi partitions that concentrates representational capacity where it is needed in a fully differentiable, end-to-end manner. Each Voronoi site carries its own embedding and migrates during training toward discriminative areas. To bridge the gap between local spatial structure and global semantic understanding, we introduce \emph{global semantic tokens}: a set of shared learnable concept tokens that distill semantic knowledge from the satellite imagery into a compact vocabulary the location encoder can reference at inference, enabling geographically distant sites covering similar environments to share semantics. TTE sets a new state of the art for location encoders across a suite of geospatial classification and regression tasks, and achieves the strongest results when used as a geographic prior for fine-grained species classification on iNaturalist-2018. Code, and weights are available at https://github.com/mvrl/TTE.

Problem

• geolocation encoders
• representational capacity
• Spherical Voronoi partitions
• semantic understanding
• geospatial representation

Innovation

• Spherical Voronoi tessellation
• learnable location encoder
• global semantic tokens
• geospatial representation learning
• differentiable geolocation encoding