Its not that silly as it seems - at least I hope so
First: good news, I fixed the socketserver issue, so that is working fine now with really big matrices as well.
http://www.unshift.de/admin/SAM/www/index.php?remoteID=468
What I wanna do:
Analyse the videoimage from a camera and extract smoothened silhouettes of people without the background. People can then make gestures to control the installation and to interact with it. This interaction is highly visual and needs a lot of CPU power, so I have to do the videoanalysis on another machine, its quite impossible to do it on one.
The imageanalysis of course works very good in EyesWeb, so i just need to send the result over to P5 - not the image, just 1/0 if theres a person.
For that I have 2 ideas:
1)Socket send all the matrix.
advantage: few CPU power on target machine required to extract the data.
disdavantage: big amount of data
2) Write an own block in EyesWeb that compressses the matrix. Since the silhouettes consist of big continuosly parts, its very easy to compress the data (count continuos same values) and send that via socket or OSC.
advantage: less data
disadvantage: more PCU power required on target machine.
Of course I can make the image much smaller and resize it in processing, that is obvious that I do not need the full resolution lateron, but I first want to shoot for the best possible result and then see whats realisitic