Mapping 2 million Pixels with CUDA

[Crazy Homeless Guy]

Not crazy far off topic, but Hayes Davidson posted this to their Twitter feed, and it was interesting in a conceptual kind of way....

 

[Sketchrender]

Not crazy far off topic, but Hayes Davidson posted this to their Twitter feed, and it was interesting in a conceptual kind of way....

 

 

Looks very cool.

but what is it proving the rendering realtime ?

I have a gtx 480 and I am sure I am not using it to its capacity at all .

rendering in realtime but with what and how.

[AJLynn]

It's using CUDA's to do a mapping of the pixels in an input video to a geometry displaced using the luminance value of the pixel in question. Which in itself isn't much of a real-world-useful thing to arch vis people but interesting nonetheless. Reminds me of things we did at Syracuse when Aaron Sprecher was teaching there. Just mapping a video to an object in 3-space isn't new (anybody else remember SGI's foray into Windows PCs with COBALT video cards - they had that running on a Pentium II, 10 or 12 years ago) but using it for parametric geometry is interesting if it works in realtime.

 

I'm not sure but I think you can do this, or something similar anyway, in Mograph. (Only you'd call it "computing" or "computation" because "compute" isn't supposed to be a noun.)