Awesome Geospatial Embeddings

This Awesome Geospatial Embeddings GitHub repository is a meticulously curated collection of academic papers. It focuses on the crucial aspect of representing Earth data—including spatial, temporal, and semantic dimensions—within embedding spaces. The repository delves into how these embeddings are defined, analyzed, and evaluated, and how they ultimately behave or are applied in various contexts.

The list specifically highlights papers that are central to the concept of geospatial embeddings, intentionally excluding generic pretraining or downstream models where embeddings are merely incidental. It serves as a vital resource for understanding the foundational principles and advanced applications of machine learning in processing and interpreting complex geographical information.

Researchers and practitioners can explore surveys, concept papers, and specific methods for location embeddings, such as those learning embeddings for coordinates, regions, or spatial contexts. The repository also clarifies that its "Dataset" column refers to publicly released embedding datasets or products, not merely datasets used for model training, making it a practical guide for accessing relevant resources.