We wanted to learn more about neural networks by making an application that recognizes hand-written digits.

What it does

It takes a user-drawn image of a digit and uses a neural network to decipher what digit it is. It then performs numerical operations.

How I built it

Using pure python - no machine learning libraries were used

Challenges I ran into

Improving digit recognition accuracy Outputting an array that the algorithm could understand Trying to make an HTML/Javscript application

Accomplishments that I'm proud of

We have zero to very little experience in neural networks and we produced a network that can recognize digits live

What I learned

Programming with neural networks Making a neural network from scratch Programming with the terminal on Linux

What's next for ArtihmeDraw

Improve accuracy by using a CNN and machine learning library Add more math functionality such that it can become a fully featured calculator with support for algebraic and analytical operations.

Share this project: