mohamedniyaz1996/web_mnist

This project allows you to scribble digits on a web UI, and have tensorflow based CNN model recognize the what you have written.

Project Title/Description

This project offers a web interface to the MNIST digit recognition model which is tensorflow based. Follow the steps to fire it up and enjoy..

Getting Started

Shows a sample implementation of how to expose a tensorflow model through a flask api and how to turn MNIST digit recognition into something fun by providing a html5 based web front-end.

Prerequisites/Dependencies

Create the conda environment with python version > 3.0

conda create --name <env_name> python=3.5
pip install keras==2.3.1
pip install tensorflow==1.15.0
pip install flask==1.1.2
pip install flask_Cors==3.0.9
pip install opencv-python==3.4.9.31

You will also need node installed. Node dependencies -

npm install connect serve-static

Firing it up

Starting the backend

./startMNISTDemo.sh

This starts up a flask app and listens in port 8080. You will have to replace keras_new with your own environment name

Starting the frontend

cd webmnist-frontend; node serve.js 

3. Open up a browser and type "http://localhost:8181" This shows up an html canvas. Draw your digit there and press submit to get the prediction out.

Author

• Pankaj Giri