anishg24 / InteractiveMNIST

Play with handwritten digit recognition on a website

Home Page:https://anishg24.github.io/InteractiveMNIST/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Interactive MNIST

This project is made by Anish Govind. Other projects can be found at my GitHub. Learn how I made this project on my Developer Blog.

GitHub followers GitHub repo size

Project Objective

Deep learning itself is very rewarding and it's great to accomplish training and testing a model. The issue now is how to allow users to play with your model so they can agree that you've done well. This is my first "full stack" ML project, incorporating Tensorflow and Tensorflow.js to create a website that predicts what digit you're writing in the 280x280 canvas space provided. It's a simple task but one that allows users to play around and see what I've made.

Methods Used

  • Inferential Statistics
  • Deep Learning
  • Convolutional Neural Networks

Technologies

  • Python 3
  • Sci-Kit Learn
  • Tensorflow
  • Tensorflow.js
  • p5.js
  • Chart.js
  • Bootstrap

Project Log

You can learn more about how I created this project and what hardships I faced in my Developer Blog.

Getting Started

  1. Click Here.
  2. Follow instructions on the site.

To-Do

  • Created a frontend
  • Process user input in the front end
  • Get predictions from user input
  • Prettify the site
  • Add a graph that shows the predictions
  • Improve model accuracy

Releases

  • 1.0.0 (6/28/2020): First working release.
  • 1.1.0 (7/3/2020): Completely redesigned site that is more user friendly.
  • 1.1.1 (7/4/2020): Fixed the user input so that the model will predict on images it's familiar with.
  • 1.2.0 (7/4/2020): Added a doughnut chart that shows prediction values.

Contributing Members

Creator: Anish Govind

Ways to contact:

IF YOU FIND ANY ISSUES OR BUGS PLEASE OPEN AN ISSUE

About

Play with handwritten digit recognition on a website

https://anishg24.github.io/InteractiveMNIST/

License:MIT License


Languages

Language:HTML 50.2%Language:JavaScript 31.4%Language:Python 15.6%Language:CSS 2.8%