See v2 branch for updated code

digit-recognizer

This predicts handwritten digits. For example, if there is a handwritten 2, it will predict it is a 2. Note that it is not perfect and will not always predict correctly.

How to use

Download all of the files Unzip numbers.zip and data.zip
Setup python (see below)
In the command line, navigate to the digit recognizer folder. Run digit_recognizer_model.py with python digit_recognizer_model.py (or just use the model that comes in the zip file) Now you can run number_recognizer.py or digit_recognizer.py which can predict user submitted numbers (examples in numbers.zip) Note that the numbers have to be put into a numbers folder

Python setup

Download the python 3 distribution from https://www.python.org/downloads/

pip install numpy
pip install tensorflow
pip install keras
pip install mlxtend
pip install opencv-python

digit_recognizer_model.py

This sets up the model using a CNN.
Note that this takes a long time to set up (it took around 15 minutes for me).

digit_recognizer.py

This only works with one digit.
This prompts the user for a filename.
The file should be in the numbers folder so that it can be found.

number_recognizer.py

This works with only two digit numbers.
This works similarly to digit_recognizer.py.
It splits it into two digits and then predicts both of the digits.
The file should be in the numbers folder so that it can be found.