Geoalert Open Urban Mapping

From OpenStreetMap Wiki
Jump to navigation Jump to search

information sign

It is an automatically updated database of building footprints that is proposed for imports.

exclamation mark

Imports and automated edits should only be carried out by those with experience and understanding of the way the OpenStreetMap community creates maps, and only with careful planning and consultation with the local community.
See Import/Guidelines and Automated Edits code of conduct for more information. Imports/automated edits which do not follow these guidelines might be reverted!

At Geoalert we apply ML/CV algorithms to Earth Observation data. In 2019 our technology prototype was applied to derive [ approx 54 mln building footprints over Russia][1]. As we continue to validate and extend auto-generated buildings coverage we look forward to contributing to the community to fill the data gaps.

Open Urban Mapping

Based in Russia, we've picked it as our testing ground. Open datasets are released by regions according to the priority of highest Geoalert / OSM count ratio.

All data is extracted from "Mapbox Satellite" since it's the only source which is explicitly permitted for processing by algorithms for non-commercial use (see license). However we look forward to providing more accurate data powered by other earth observation sources.

Data download

Download datasets from our project's Github page.

As Mapbox satellite has comparatively poor quality for Russian territory (too old, a lot of snow, clouds, only panchrome etc.), the data should be validated quite carefully and updated based on some premium imagery available for the manual mapping.

Mapping contribution

You can use this data to contribute to OSM and accelerate your mapping process. Since the data is generated automatically it shouldn't be imported before you check and validate. You can preview this data using Quantum GIS, Openstreetmap ID ("custom Map Data"), whatever you want, and edit it locally (JOSM etc.)

Data accuracy

Our metrics show that in the vast majority of cases the quality is at least as good as hand digitized buildings in OpenStreetMap. It is not perfect, particularly in dense urban areas but it is still awesome


Yes, we can call after Microsoft about our metrics and our belief in our data awesomeness. But... really. We aim to provide sustainable segmentation output even if the input imagery is not so good as MS or Facebook may have.*kyvtjKx3yvftW3UjuSVn0Q.png

The more detailed description will be provided as we get through more experiments.


The "Open Urban Mapping" project data is licensed under the Open Database License (ODbL) to be compatible with OSM. The input data is copyrighted by the data providers. The Mapbox Terms of Service states that

You may use Studio or third-party software to trace Mapbox maps solely comprised of satellite imagery ("Mapbox Satellite Imagery") and produce derivative vector datasets for non-commercial purposes and for OpenStreetMap. ([YmuSEapb])

Technical details

  • Coordinate reference system - EPSG: 4326
  • Data format - GeoPackage or GeoJSON

Changeset tags

If you edit and import this data to OSM, please refer to

source = Mapbox/Geoalert