[robocup-small] SSL-Vision plugin for semi-automatic color calibration - beta

[Angelo Gurzoni Jr]

Hello everybody

At last I could find some free time to finish the porting of a
semi-automatic color calibration algorithm as a SSL-Vision plug-in. It's
based on a MLP Neural Network, a sort of commonly used solution for color
calibration.

Would be great to know if works well and if it's useful. If there's demand,
then we can incorporate this on the source code's repository. 

Please take a look and share feedback. I uploaded a tar.gz file with the
source on the downloads section of the SSL-Vision project at code.google.com
(also feel free to email me if you don't have access to this website and
I'll email the file)

 

This new and the existing, manual color calibration plug-in, can coexist
side-by-side without any trouble. The only requirement added by the new
plug-in is the presence of the OpenCV 1.0 ML library.

The usage is simple:

(i)                  capture some examples of each color, 4 or 5 samples are
enough 

(ii)                click train network

(iii)               click load neural to LUT

 

 

** Tks to Danial Esmaeli for providing a SOM network code, which I still
intend to port and try it for a fully auto calibration :-) 

 

 

Regards to you all

Angelo

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-small/attachments/20120223/f6679aa5/attachment.html>