ann

BP Artificial Neural Network with three layer.
Using this network to recognize MNIST handwritten digits.

require library

1. python2.7
2. numpy

   pip install numpy

3. pillow

   pip install pillow

4. GUI need PyQt4

https://www.riverbankcomputing.com/software/pyqt/download

you can install from binary package here PyQt4-4.11.4-gpl-Py2.7-Qt4.8.7-x32.exe

1. Mnist Database

http://yann.lecun.com/exdb/mnist/

include

• train-labels-idx1-ubyte
• train-images-idx3-ubyte
• t10k-labels-idx1-ubyte
• t10k-images-idx3-ubyte

get started

1. copy data.json from best dictionary to root dictionary(with file_tools.py same dictionary).
2. copy t10k-images.idx3-ubyte file to img dictionary and run img.py (ensure PIL installed) to get MNIST images.
3. run gui.pyw (ensure PyQt4 installed),you can drag MNIST image to GUI program.
4. enjoy and happy coding.

screenshot