Early this year Microsoft released approximately 125 million building footprint polygon geometries in all 50 US States in GeoJSON format. This dataset containing 125,192,184 computer generated building footprints was the output of a combination of deep learning, computer vision and artificial intelligence work by Microsoft’s Bing Maps Team.
The team has reportedly been been applying these techniques with the goal to increase the coverage of building footprints available for OpenStreetMap. Before the release there was an estimated 30,567,953 building footprints in the US on OpenStreetMap.
Deep learning, computer vision and artificial intelligence was applied to generate the data
The Bing Maps team used Open Source CNTK Unified Toolkit which was developed by Microsoft and applied Deep Neural Networks and the ResNet34 with RefineNet up-sampling layers to detect building footprints from the Bing imagery.
To remove noise and suspicious data (false positives) from the predictions the team applied a polygonization algorithm to detect building edges and angles to create a proper building footprint.
You can read more about this work and directly download the data from GitHub. The CNTK toolkit developed by Microsoft is open source and available on GitHub as well. The ResNet3 model is open source and available on GitHub.
Also read read this post on how to apply machine learning in remote sensing.