Processing the audio is done by autocorrelation, a very simple technique explained on Alex Ellis' blog. The main limitation is that it can't properly detect chords, only single notes at a time. That's ok, I'm still learning how to read one note at a time anyway.
Displaying the notes is done by tranform: translate(x, y) SVG path
elements from JavaScript. That was a bad idea.
I took most of the SVG path from [Wikipedia's music symbol] (https://en.wikipedia.org/wiki/List_of_musical_symbols).