On this site you can draw a digit and the machine learing algorithm will guess what you have drawn. The app is deployed with Heroku and is available here! However you can also run it locally by following these steps:

1. Firstly, clone the repository: git clone https://github.com/thdb-theo/doodle-guessing/
2. Then, locate the directory and the start the server with npm run develop
3. Lastly, enter the directory src/api and the start the backend server with python listener.py. Make sure you have installed all the necessary requirements.
4. Now you can go to localhost:8000 in your favourite webbrowser and start drawing!

• npm
• tensorflow & keras
• scikit-image
• numpy

The classifier uses a machine learning algorithm called a Convolutional Neural Network (CNN). The model has been trained using the MNIST database of 60000 handwritten digits.

When you draw a digit, we send the data to a separate server using a HTTPS requst. The server is written in Python which parses the data and transforms the drawing to a matrix. Unfortunately, the size of the canvas (400x400) and the size of the training images (28x28) is vastly different. To scale down the image, we use bilinear intorpolation. We send the scaled down image through the classifier which guesses what digit was drawn.

Hooray! It guessed correctly! 🎉 🎉

However, noone is perfect 🤷