Locamage

The universal geospatial model

Ask any place what changed.

Point Locamage at imagery from any satellite, plane, drone, or street camera and ask in the words you already use. It reads the scene and answers.

Reads optical

Sentinel-2 harvest composite of Iowa farmland with every field boundary predicted by Locamage outlined in blue
Sentinel-2Iowa · Oct 2024
10 m/px · field boundaries

Where does each field end?

Every parcel in the scene is outlined from a single October pass, harvested and standing fields alike. The boundaries follow the real edge of each field, not the survey grid, and carry over to the next revisit.

Already reading imagery from every major feed

What it does

One model answers every question, from every source.

Map, monitor, find, compare, measure. The same model handles all of it, whatever satellite, plane, or drone took the picture.

01Map

What is on this land?

It draws the map from the pixels up. Every rooftop, road, field, and pond, labeled and measured, ready to check against a permit or a claim.

  • optical
  • radar
  • elevation
Aerial view of Lohur with every building picked out in blue

Map

Land cover · Buildings · Roads

02Monitor

Are the crops okay?

It reads every field the way an agronomist would, all of them at once, week after week. Stress and patchy growth get flagged while there is still time to act.

  • optical
  • hyperspectral
  • revisits
Aerial view of farmland around Aachen with agricultural parcels marked in blue

Monitor

Crops · Forests · Water

03Find

Where are all the ponds?

Describe what you are hunting for, ponds, solar panels, new construction, and it checks every acre in the region, then marks the hits.

  • optical
  • elevation
Aerial view of Dhaka with every pond marked in blue

Find

Water · Solar · Construction

04Compare

What changed since March?

Give it two dates, or every pass in the archive, and it outlines what actually moved. Clouds and shifting light get ignored.

  • optical
  • radar
  • revisits
Two-date Sentinel-2 city scene with detected changes highlighted in blue

Compare

Two dates · Whole archive

05Measure

How much forest came back?

Some questions end in a number. It measures from orbit and shows you where the number came from.

  • optical
  • hyperspectral
  • elevation
Aerial view near Dortmund with tree cover marked in blue

Measure

Acres · Biomass · Extent

The model

One model. Any sensor.

Most Earth AI is built for one satellite and falls apart on the next. Locamage was built the other way around.

Sentinel-2 harvest composite of Iowa farmland used to illustrate the model reading many sensors at once

One model reads them together.

Plain words: where to look, what changed, and what to do about it.

  • 01Opticalstreet level to whole-country mosaics
  • 02Radarstructure through cloud and darkness
  • 03Hyperspectralhundreds of bands past visible light
  • 04Elevationterrain read as part of the scene
  • 05Revisitsthe same acre, week after week
  • 06Unseen sensorsnew feeds read on arrival, no retraining

Thin-map places read clearly

It learned from imagery covering the whole planet, so an unsurveyed region reads as well as a mapped one.

Your vocabulary, not ours

Mark a few examples of the thing you track and it maps the rest using your labels.

Your archive becomes searchable

The imagery you already pay to store turns into a queryable history of any acre.

Request access

Somewhere in orbit there is a picture of your problem.

Tell us what imagery you have and what you need to know from it. You will get a straight answer about whether Locamage fits.