This is a project which I used to learn machine learning. Here I use support vector machines to learn from an existing dataset of handwritten digits in the form of 8x8 pixel images.
This is an image of the number 0 in an 8x8 pixel grid.
This particular image is converted to an array of numbers which makes it possible to process. A white square is represented as 0 and a black square is represented as 15. Here's what the 0 is converted to:
The SVM learns from multiple such images and you train the model to recognize digits based on these values.