SAM 2

Segments objects in images and video without task-specific training.

Field
Cross-domain
Method
Foundation model
Data it takes
imagery
License
Apache-2.0 AND BSD-3-Clause

SAM 2 outlines objects in images and video given a point, a box, or no prompt at all, without task-specific training. Paired with samgeo, masks are returned georeferenced when the input raster carries a coordinate system.

It is the general-purpose starting point where no specialized model exists for the subject, and applies to aerial, satellite, drone, handheld, camera-trap and microscope imagery.

Upstream project →

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

Run SAM 2 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