I have developed a machine learning architecture that suggests missing road surface tags from overhead imagery. It is currently implemented using USGS NAIP imagery which has coverage for the US. The goal of this project is to provide a dataset that can assist OSM-based routers in places where road surface data is unknown (e.g. TIGER import in rural areas). This should allow OSM-based routers to make optimal decisions in these areas and improve travel time estimation.
No recording is available from this session.