Deep learning and street imagery based methods serve as a very good complement to the traditional survey-based and administrative data sources for more comprehensive urban surveillance, thus helping to achieve a more balanced development with finer granularity.