This AI Python Application can identify any number you draw!
It works by training a CNN (Convolutional Neural Network) on thousands of handwritten digits. Through this process, the model learns the underlying patterns behind them and creates the weights necessary to recognize new digits. To make training faster, I added GPU support, and implemented save/load functionality so the model doesn’t need to be retrained each time.
Before the model can make predictions, the drawing needs to get preprocessed by converting it to grayscale, centering, resizing, and pixelating the image. Once that’s done, it can identify any digit almost instantly.
I used PyTorch for the model and training, PIL for image preprocessing, and Tkinter/matplotlib for the UI and visualization.
This was a fun project that gave me great hands-on experience with Machine Learning, GPU acceleration, and Image Processing.
Github Link:
https://github.com/NicholasMalamud/AI-Number-Recognizer