| Since growing up of recent parallel computing facilities, we have assisted to a development of visualization programs to successfully trace down the output more and more complex. Graphic output is the fastest way to show and probably understand large amount of numbers, see their properties, desities and so over. However visualizations programs are becoming more and more complex and hard-computable, seldomly a dedicated machine can be used for computing. Therefore, using distribuited computing is the best way to solve the computing overhead due to graphics interfaces. In this environment we developed our project. |
Our project was :
CNN is one of most used tools in Image Processing.
We inserted the CNN code in an AVS
Module
.
The result is a generic module for image processing with many applications
available.
Indeed, modifing parameters of both aCNN and
GenMask (a 3x3 filter generator) this module
offers many different application on images.
Many books are available for a detailed description of which parameters
are appliable into a CNN.
A library of applyable filters is under construction. It simply shows
a list of names (or their results) and, if selected, automatically sets
parameters.
We'll never go through it, unless someone is really interested.
This module is available at IAC
(International AVS Center)
or it will be available (if someone needs it) at FTP
site in Italy
See also : Neural Networks Links See also : AVS
generalities
We made a module to assert the possibility to use PVM libraries inside
an AVS module. This module is not currently available for its rough programming
style, it is indeed a proof that it can be done quite easily, the program
itself was *not* neat. We used simple embedded PVM instructions into a
standard AVS module.
Our program reads a gif image from a distribuited process on a remote machine.
Then sends it, through standard PVM instructions. AVS module reads it using
PVM commands and then shows it in an AVS window.
|
