Slashdot Mirror
Optical Camouflage Puts Kinect Into Stealth Mode
UgLyPuNk writes "Takayuki Fukatsu, a Japanese coder who works under the name Art & Mobile, has done a bit of trickery with Kinect and openFrameworks. The peripheral will still track your movement and position, but turns your image nearly transparent. Take a look (it's particularly obvious at about 1:30):"
I think it very much more simple than all of that, he's obviously just taken a couple of polaroids stuck them in front of the cameras and then he's just walking around in front of the depth sensor, so the software is just using the depth information to alter the still images. Not really clever at all and as far as I can see damn obvious from the video (the pink border which shows on the right hand side).
Notice how when he walks far back, the distortion goes away and he disappears completely. What does that tell you?
So, he is using depth info, but what he is doing with it is rather lame. He still has a static image of the empty room, otherwise when the person went far enough back, he would have to appear uncloaked not disappear completely. Of course you would need 2 kinects and much more work to avoid the need for a static background image and just apply the "cloak" to objects nearer than the background. But that would certainly be cool.
What we have here, you can do better without a kinect by simple diffing of the background image. If he at least used the depth info to alter the distortion it would be interesting, but it seems to me that when he is walking towards or away from the camera the distortion does not change at all.
Violence is the last refuge of the incompetent. Polar Scope Align for iOS
"pre existing image of the background" this makes it an epic fail, but cute project.
I'l be impressed if it generates the background without ANY reference images or reference data.
Do not look at laser with remaining good eye.