Computer vision project reconstructs 3D cities/landmarks in hours using Flickr photos

they generated a skeleton of Rome in 21 hours using 500 compute cores and 150,000 photos