Hallo!
I am trying to resize a live video feed to a very small picture (16x12), extract a list of rgba values from this image and pass the list to ann_mlp.
just an idea: If you want to process the pixel data in a neural net I think it would be much faster if you write a GEM-external and process the data in C++ with fann (I did something similar a while ago ...), because maybe you'll need a lot of CPU :)
LG Georg