Old news, but still, very cool technology. If I had the money, I would want to do a PhD on how this technology could be used to start actually <i>understanding</i> the scene - I.e. try to segment out items in the scene, estimate physical properties, then use physics models to try to explore affordances, etc... The goal being to go from a series of images of a scene to a set of annotated objects that could be used by a planning system. No so much "What can I see here" as "What can I do with what I see in front of me". An artificial MacGyver of a sort.