In this particular form, it's probably not very useful at all -- but identifying edges, or extracting 'foreground' images from 'background' images automatically with clean edges is something that would be ridiculously useful for videography, photo editing, or just graphic design in general.

Furthermore, if you advance the technology to being able to scan a photo of, say, an airport lobby, break each component into layers, and then perform shape identification on each of the identified pieces, you've got left-package detection or weapon identification technologies available to you (potentially.)