mnist-draw

This repository contains a single page website that enables users to hand-draw and classify digits (0-9) using machine learning. A machine learning model trained against the MNIST dataset is used for classification.

Setup

Python 3.5+ is required for compatability with all required modules

# Clone this repository
git clone https://github.com/rhammell/mnist-draw.git

# Go into the repository
cd mnist-draw

# Install required modules
pip install -r requirements.txt

Usage

To launch the website, begin by starting a Python server from the repository folder:

# Start Python server
python -m http.server --cgi 8000

Then open a browser and navigate to http://localhost:8000/index.html to view it.

An example of the website's interface is shown below. Users are guided to draw a digit (0-9) on the empty canvas and then hit the 'Predict' button to process their drawing. Allow up to 1 minute for the processing to complete. Any errors during processing will be indicated with a warning icon and printed to the console.

Results are displayed as a bar graph where each classification label recieves a score between 0.0 and 1.0 from the machine learning model. Clear the canvas with the 'Clear' button to draw and process other digits.

Interface example:

Machine Learning Model

Python scripts related to defining, training, and implementing the machine learning model are contained within the cgi-bin folder.

A convolutional neural network (CNN) is defined within the model.py module using the TFLearn library. This model is configured for MNIST data inputs.

The defined CNN can be trained against the MNIST dataset by running the train.py script. This script will automaticallly load the MNIST dataset from the TFLearn library to use as input, and the trained model's parameter files are saved into the models directory. Pre-trained model files are made available in this directory already.

The mnist.py script implements this trained model against the user's hand-drawn input. When the 'Predict' button is clicked, the contents of the drawing canvas are posted to this script as data url, and a JSON object containing the model's predictions is returned.