From the Google Earth blog:
When discussing the new Ovi Maps 3D a few days ago, one interesting point about their maps are how they are generated -- hundreds of aerial photographs, automatically converted to 3D.
It's apparently growing into a popular technique, as Pix4D is showing off some similar technologies.
In the article they mention using the Swinglet CAM as a great way to capture the imagery for uses like this. The "4D" part of the equation is time; you can view the progress of 3D buildings through time, assuming you've done enough captures of a particular area.