Bing Maps may not be the default geo-locational resource most people use in a pinch, but when it comes to research solutions for the open data community it's becoming harder to deny the advancements in computer vision and geospatial intelligence the Bing Maps team are achieving.
Recently, Bing Maps collaborated with Statistics Canada to provide a 12 million building footprint map to the Open Data community that include visual mapping of buildings across all Canadian Provinces and Territories.
The free-to-download resources can be snagged over at its GitHub repository, where Microsoft explains in great detail of the process taken to deliver over 12,663,475 footprints.
What the data include:
12,663,475 building footprint polygon geometries from all Canadian provinces and territories in GeoJSON format.
A summary review of the process reveals that Microsoft had to identify the difference in biomes and terrains between Bing Maps typical US-based mapping and that of the Canadian landscape. Next Microsoft implemented a segmentation-based solution built on CNTK to optimize its overlay's of different data sets.
The FAQ over at GitHub is rather detailed for anyone interested in understanding GeoJSON, Semantic Segmentation, Polygonization, coordinated reference systems or just more about OpenStreetMap.