Clay

Creates embeddings from satellite imagery that can be used for classification and other downstream tasks with relatively few labeled examples.

Field
Remote sensing
Method
Foundation model
Data it takes
imagery
License
Apache-2.0

Clay is a vision transformer pretrained across Sentinel-2, Sentinel-1, Landsat, NAIP and MODIS. Rather than answering a fixed question, it converts imagery into embeddings: numerical representations of what a patch of ground looks like.

The practical effect is a reduced labeling requirement. A scene is embedded once, then a classifier is fitted on a few dozen labeled examples rather than thousands. This suits projects needing a custom classification without the budget to build a large training set.

Upstream project →

Catalog entry last checked 2026-08-17. All models →

Run Clay 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