Very cool! Can we tweak your algorithm a little to be seeded with a real photo from a single viewpoint?