Depth Anything 3

Recovers depth, multi-view geometry and camera pose from ordinary photographs.

Field
Cross-domain
Method
Foundation model
Data it takes
imagery
License
Apache-2.0

A single model covers three tasks that usually need three: depth from one image, geometry from several, and the camera poses that relate them.

Depth from a single photograph is relative. Converting it to metres requires a known dimension in the scene or a second view.

Both the code and the weights are Apache-2.0, which is why it supersedes Depth-Anything-V2 and VGGT in the catalog — those carry non-commercial weights.

Upstream project → · Paper →

Catalog entry last checked 2026-09-08. All models →

Run Depth Anything 3 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