SatCLIP
Turns a longitude and latitude into an embedding that describes the place, for use as a covariate.
- Field
- Cross-domain
- Method
- Foundation model
- Data it takes
- location
- License
- MIT
SatCLIP maps a longitude and latitude to a 256-dimensional embedding learned by pairing locations with Sentinel-2 imagery. The embedding stands in for what is known about a place, as an input to geospatial regression or classification.
It captures landscape-scale structure and separates nearby locations poorly. GeoCLIP works in the opposite direction, from a photograph to a location.
The code and all six checkpoints are MIT-licensed; upstream scopes their use to research.
Catalog entry last checked 2026-10-03. All models →
Run SatCLIP on your data
The lab runs this on its own compute, in a container, with the inputs and parameters recorded alongside the result. Initial scoping conversations are free.
Contact the Lab