How to write the code for such 4X12 Image

I am working in bank note (using pattern recognition), and I would like to training my neural network. by now I have divide each bank note for 4X12 split (parts) and I found the average for each parts. I have done the following step and I would like to know if I'm going on the right way or not and if there is any suggestion:

1\ I converted the 6 bank note to gray scale. 2\ I split each bank note to 4X12 parts. 3\ I calculated the average for each bank note image parts. 4\ I used bank note (head upright, head reversed, tail upright, tail reversed) for each bank note.

so would you help my in the matlab code for the input, traing the neural network and target, looking forward to hear from you soon, thanks.

this was the answer before from Mr. Greg Heath, thankx to him Single inputs to the net must be vectors of the same size. Therefore columnize your images using the (:) operator. Your input and target matrices should have the dimensions

[ I N ] = [ 48 24 ] = size(inputs)

[ O N ] = [ 6 24 ] = size(targets)

You can rearrange your 24 I/O column pairs anyway you want. For example,

targets = repmat(eye(6),1,4)

But still I'm looking for how to write the neural network (input values, target values, and output). How many layer will be create it and what are the codes (learning,training, comparing function of inputs & targets ), would you please give me general Idea or any such system similar to this, looking forward to hear from you soon.

Answer 1

[Greg Heath]

If you search using

greg patternnet

I'm sure you can find useful examples.

Hope this helps

Greg