University of Washington researchers have created the first publicly available dataset of labeled classical music, named MusicNet. "MusicNet is a collection of 330 freely-licensed classical music recordings, together with over 1 million annotated labels indicating the precise time of each note every recording, the instrument that plays each note, and the note's position in the metrical structure of the composition," described in its homepage. It aims to explore machine learning techniques in the realm of music, including identifying the notes performed at specific times in a recording, classifying the instruments that perform in a recording, classifying the composer of a recording, identifying precise onset times of the notes in a recording, and predicting the next note in a recording, conditioned on history. For more information, please visit MusicNet homepage and the team's paper "Learning Features of Music from Scratch".

Courtesy of MusicNet.