I am currently collecting verified existing land use data for major US cities, I will combine them with urban remote sensing data, and publish a labeled training set. I wish the set can help planners and geologists to develop more accurate computer vision algorithms for classifying land use.