There's a couple source separation algorithms that do a pretty good job splitting the audio into single instrument tracks (I tried open_unmix) -- I'm struggling to transform the single instrument audio into frequencies or notes from that one instrument track.