How do we build geospatial machine learning that works reliably across views, resolutions, and sensors worldwide? Geospatial AI is our platform for planet-scale understanding: unified embeddings, cross-view localization, retrieval, and structured reasoning over satellite, aerial, and ground imagery. Recent work includes retrieval-augmented neural fields for multi-resolution geo-embeddings (RANGE), Earth Embeddings for AI-centric planetary representations, goal modality agnostic active geo-localization (GOMAA-Geo), and fine-grained satellite image synthesis with structured semantics (VectorSynth). We integrate diverse geospatial data sources to support environmental monitoring, agriculture, transportation safety, and conservation applications described in our ecology and remote sensing areas.
Spotlight publications tagged for this research area.