What tool/library can help me to generate and test model of neural network that would work with X*Y int arrays as input and N*4 int arrays as output?

I want to make simple neural network that from given gray-scale 2d arrays like

[[1 1 1]
 [0 0 0]
 [1 1 1]] 

would give me line coordinates

[[1 2]
 [3 2]]

I have a C++ program that has a real algorithm in it to compute lines from gray-scale images and a large set of images. I need some NN tool wit scripting language so that I could generate a program to teach and generate NN. so I could create some programm with structure like:

generateNN(NN);
NN.provideData(WxHGSImageArray);
NN.provideDesiredResult(2PointsLineArray);

So... Is there any such C++/C开发者_开发知识库 lib for generation of NNs or any programmable tool with scripting language?

Main problem here is - what tool/lib supports creation of systems that can take X*Y arrays and output N*4 arrays (N - number of lines found on image)? (btw X and Y are different for different images)

Try SNNS. It is a great software package for simulating neural networks. A JNNS version of the toolsuite is also available. It is called JavaNNS.

I've used the SNNS and JavaNNS for neural networks simulations and it was very useful. It has a steep learning curve, but it's certainly worth trying!

Here is a quite comprehensive list of nn simulators: Comparison of Neural Network Simulators

If you already have a C++ neural network library you're comfortable with, you might try writing a small binding for lua (take a look at toalua++ in this case) or Python (SWIG is a good option).

Of course, if the API is too big with might not be worth the effort.