Hacker News Reader

From what I gather there is a number of algorithms that work pretty well, but only on certain datasets and for certain parts of the world. For example, one approach could be flawless to extract highways and on and off ramps in the US, but the same algorithm may struggle with French roads.

What I am excited about within the context of open data and crowdsourcing is using machines to do a rough first pass over the imagery, and get people to rate or fix the results.

A wonderful example of this is the New York Public Library's Building Inspector: http://buildinginspector.nypl.org/

There's various algorithm in use for OSM, eg click in the middle of a lake and it'll trace the outline for you based on color, but going from a satellite image to a map requires both a human and other sources of info, generally.

Are you sure they don't already do this, and just only use it on Google Maps?