Hi everyone! It's been a while hasn't it? :) I recently graduated from university, specializing in Computer Science and Artificial Intelligence. Since A.I. is all the rage these days, I though it would be fun to celebrate by sharing some projects related to this fascinating field of study. This project contains a "simple" A.I. / machine learning model (neural network) to recognize handwritten digits! Try it out by painting a digit (hold down left mouse button while drawing) and hit space to let the Scratch Cat guess what you have drawn! (This project was originally shared on my test account, but I've improved it and thought it was worth sharing here too :))
Works best if you draw the digit centered and with roughly the same size as the grey circle. --> UPDATE: the drawing will now automatically be centered! Should improve the accuracy :) It's not perfect though, so it will not get it right 100% of the time! Music: Vexento - Wind Chimes https://soundcloud.com/vexento/wind-chimes Want to understand how this works? This wonderful youtube series by 3blue1brown is a great starting point (warning! maths :)) https://www.youtube.com/watch?v=aircAruvnKk&list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi The model was trained on the MNIST handwritten digits dataset. http://yann.lecun.com/exdb/mnist/ For those interested, the training code (Python, TensorFlow) can be found here: https://colab.research.google.com/drive/1mC4fszCTYOw5CkfinJs9d41riOPBqNpx?usp=sharing
