DINOv3-SAT

Extracts dense visual features from aerial and satellite RGB imagery, used as a frozen backbone for downstream tasks.

Field
Remote sensing
Method
Foundation model
Data it takes
imagery
License
LicenseRef-DINOv3 (not open source) — restricted use

DINOv3-SAT is a ViT-L/16 backbone trained on 493 million satellite images. It produces dense per-patch features for RGB aerial and satellite orthoimagery.

It is an encoder only, with no task head. A classifier or segmentation head is trained on the features it produces, and that head belongs to whoever trains it.

The weights come under Meta’s own license rather than an open-source one. Publications reporting results must acknowledge DINO Materials.

Upstream project →

Catalog entry last checked 2026-08-18. All models →

Run DINOv3-SAT 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