CUDA is amazing.
This technology is leading super computing scene today.
Before talk about Kimura's sophisticated implementation,
I'd like to start from more physical, basic one.
If you want CUDA working, you can buy one of CUDA enabled video card.
If you are eager enough to seek nearly high end with easy installation.
You can chose reasonable priced quadro 4000.
This model is selling under $1000 with 350+ GFLOPS (Double Precision, if single,this hits 700 GFLOPS).
Another Cheaper but newer model named Quadro 2000 is available with 60% price of 4000 and 80% performance.
If you use CUDA on adobe After Effects CS5 (latest on 2010 Dec),
adobe said that OSX environment is a bit faster than Windows 7 (64).
Ok, back to the main issue.
This Kimura's realtime visual attention estimation from video clip is amazing!!
composed of a dynamic Bayesian network with four layers that combines several fundamental statistical models. The proposed model enable us to
-automatically estimate eye focusing positions an their densities only from video frames,
-automatically derive model parameters of the network with the EM algorithm when eye tracking sequences of human subjects are available,
-execute the estimation procedure in near real-time (70 msec/frame @ 640x480 pixels) with stream processing through such as GPUs, by introducing MCMC-based particle filter.
watch the demo video how it works and be amazed.
If I am a bio-logging guy, I'll use this technique to see what is the animal's focus point on their view.
Bio Logging (in Japanese)
http://bre.soc.i.kyoto-u.ac.jp/bls/index.php?%BF%B7%A4%B7%A4%A4%C8%AF%B8%AB
