Information from Microsoft
The Microsoft Maps Team has been leveraging the company’s investment in deep learning, computer vision, and AI to identify and map features at scale and produce high-quality building footprint datasets. It plans is to add the datasets to the OpenStreetMap and MissingMaps humanitarian efforts.
The following locations are available and Microsoft offers access to this data under the Open Data Commons Open Database License (ODbL).
| Country/region | Million buildings |
| Nigeria and Kenya | 50.5 |
| Uganda and Tanzania | 17.9 |
| United States of America | 129.6 |
| South America | 44.5 |
| Canada | 11.8 |
| Australia | 11.3 |
The vintage of the footprints depends on the collection date of the underlying imagery. Bing Maps Imagery is a composite of multiple sources with different capture dates (ranging 2012 to 2021). Each footprint has a capture date tag associated, as deduced from the imagery used.
The building extraction is done in two stages, and uses Microsoft’s Open Source CNTK Unified Toolkit:
- Semantic segmentation – Recognising building pixels on the satellite image using deep neural networks
- Polygonisation – Converting building pixel blobs into polygons
The building footprint polygon geometries are then made available in line-delimited GeoJSON format for download on Github.
Quality metrics show that in the vast majority of cases the quality of the datasets is at least as good as hand digitised buildings in OpenStreetMap. It is not perfect, particularly in dense urban areas, but it provides good recall in rural areas. Read more about Microsoft Maps Open Source projects.
