Numeric Digit Recognition

This is an interesting, if contrived, example of how artificial neural networks allow computers to do character recognition. Put simply, this project trains a neural network to recognise the digits 0-9 from handwritten input.

See the network categorize digits here.

The source code for this example is here.

The data used for training and testing is contained in the optdigits.tra (training) and optdigits.tes (testing) files. The optdigits.names file contains a description of the size and shape of the data. The source of this data is the Optical Recognition of Handwritten Digits, a Creative Commons licensed dataset containing 5620 images pre-organised into training and testing sets.

Just run the train.py script to emit a JSON representation of the trained neural network (saved as the file, nn.json).