How do we extract reliable structure and semantics from overhead imagery across resolutions and landscapes? We develop remote sensing and mapping methods for satellite and aerial data—segmentation, synthesis, atmospheric inference, and large-scale geospatial products. Recent work includes retrieval-augmented neural fields for multi-resolution geo-embeddings (RANGE), fine-grained satellite image synthesis with structured semantics (VectorSynth), global agricultural field boundary delineation (Fields of the World), and deep learning for particulate matter and land-cover mapping. This area emphasizes sensing modalities and map-level outputs; biodiversity science outcomes are described in our ecology area.
Spotlight publications tagged for this research area.