Oceanir / Physical inference

MODELManta, a world-class geospatial model

Physical inference for the world.

Oceanir uses evidence to work out what’s true about the physical world. We’re starting with geolocation and building toward understanding places and events from photos, video, and live sensor feeds.

Try geolocation free
Manta locating a photo: a parking-lot photo is dropped into the workspace, placed on the map, resolved to a street address and checked against Street View
01 / Geolocation · Recorded workflow

See why a location was suggested.

Check the clues, compare possible locations, and save your review. The estimate gives you a place to start, but it still needs checking.

Understanding
starts with where.

Knowing where an observation belongs connects it to the world around it. Find a place, examine its surroundings, and check a claim against the evidence, all in one workspace.

Oceanir · Physical inference

The world leaves
evidence.

A photograph.
A video frame.
An observation.

An aerial photograph of streets and buildings

01 Evidence

Orca · Reading the scene

Find the signals.
Connect the context.

02 Observe

Physical inference

Architecture

Building forms. Materials. The details of a place.

03 Reason

Oceanir · A closer look
Roofs and street layouts as visual clues

Do the details
line up with
the place?

Compare the candidates.
Follow the evidence.
Keep the uncertainty.

04 Review

05 Explore

Evidence → location → understanding.
In the workspaceLocate · Understand · Verify
Oceanir / Observation → analysis → reviewInterface preview
CANDIDATE PLACESKYLINESIGNAGEVEGETATIONARCHITECTUREROAD MARKINGSTERRAINFOUNDSTILL OPEN
GeolocationExample
A hillside skyline with a domed building and minarets Roofline & skyline

Start with the scene

A dome and minaret profiles are visible above the surrounding buildings.

Observation
Added
Location estimate
Run analysis
Next step
Inspect candidates
Run a location analysis

Illustrative observations. Open the workspace to run an actual analysis.

01 / Geolocation

Place an observation in the world.

Start with geolocation. Inspect candidate places and the visible clues that point to them.

Explore the workflow

02 / Who it’s for

Understanding for people. A foundation for machines.

Answers you can examine.

Investigate a place or event and see the evidence behind the answer. Check the reasoning, consider what’s uncertain, and keep a record for later review.

Caspar David Friedrich, 1818

Know where you are. Understand what’s around you.

We’re also building this for robots, vehicles, and other embodied AI systems that need to locate themselves and understand their surroundings.

J. M. W. Turner, 1844

Our goal

One model. Any place on Earth. Any trace of it.

One model that understands the physical world from photos, video, and live sensor feeds, for the people and machines that need to act on it.

Albert Bierstadt, 1868

Questions

The common ones are here. For anything else, ask Oceanir below.

Talk to us

Answers use Oceanir’s product information. Please keep personal details out of your question.