Use Machine Learning to Create Building Heights in OSM

Byron North
Friday, 10:20am – 20 min
Chris Helm, Pramukta Kumar and Sean Gorman
DigitalGlobe

In this talk we’ll discuss how we used OSM building footprints and height data to validate building heights extracted from satellite imagery. Then how this validated data set was used to build a model for generating building heights for OSM data in which the attribute is missing.

This approach will be used as an exemplar of how OSM can be leveraged as training data for solving a host of object detection challenges - e.g. generating building footprints, enhanced post disaster change detection.

Lastly, we will touch on the pros and cons of using automated models for potential OSM data creation and contribution.