Greg Borenstein

Timo -- You might also really like Adam Harvey's videos that visualize OpenCV routines, for example this one about face detection: vimeo.com/12774628

I keep thinking about whether these are equivalent to the Terminator HUDs that Slavin mocks (images.wikia.com/terminator/images/2/25/T-800a_Threat.jpg). Why would a computer communicate to itself with text? These visualizations are really for the human observer of the CV process. They're akin to Rodney Brooks's idea of language having been invented by god to make it easier to read our minds. In this case these graphics give a window on the extent to which the CV algorithms are seeing the world the way we want them to, whether their vision agrees with ours. It's not an internal representation, it's a performance for our benefit. Like Kyle, find myself doing a dance of rapidly connecting the displays with the semantics of what they represent: arrows for flow, boxes for blobs, etc. These graphics were designed by people to be seen by other people. They're meant to let us see how the algorithm is doing. It's only an internal state of the computer to the extent to which conversing is mind reading.

Ha! Thanks. Yeah, it's pretty fun. I want to do it at scale so that it can actually aim something at you. Actually just had a conversation about possibly building a giant one for next NY MakerFaire...