||Topic: optical character recognition
||Posted: 31-Oct-06 05:51
i'm new at neural networks.
i went through many neural network tutorials.for back propagating multi layer neural networks.
i am trying to write an optical recognition software for fun.
i'm trying to recognize the 26 alaphabatic English characters.
lets say my character image is 80*40 in pixel values.
1.so how can i feed the neural network inputs?(how many neurons)
obviously i can't send one pixel value,
2.how much hidden nodes should i use?
assuming that there are 7 output neurons,that represent the ASCII value of the character.