DigitRecogniser

A website to detect Handwritten digits !

The tech stack used

HTML5, CSS3, JS (Es-6), NodeJS, REST-style API, Express, Python, Scikit-Learn, Numpy, Pandas, OpenCV, MNIST Dataset

Running

1. Fork and clone the repository
2. Install NodeJS
3. Run command on terminal: node server.js
4. On you favourite browser go to: https://localhost:3000

Features

Technical Features

• The information about the pixel data is transmitted to backend NodeJS.
• NodeJS then runs a terminal script.
• This run a Python file which uses a pretrained neural network to get the predicted data.
• The number then travels back to the NodeJS server an d is displayed on the screen
• There is a also a latency calculator which determines the time taken for the entire process.
• The average time is usually around 0.9s.
• The canvas is made of 28 x 28 pixels.

Miscellaneous

Video Description: https://drive.google.com/drive/folders/1PcGVGBMJYeitez-ZzJeVxag2czqqEqDJ?usp=share_link

Click on @nightfalcon to know more ;)