I’m interested in the China/global economy and this is one of the many data points.
Are you doing some kind of manual sampling or have you automated parts of it? Also, what kinds of objects can be distinguished with the images you use? I assume it might require a bit of practice to reliably recognize certain kinds of structures.