POP909 Music Dataset

The POP909 dataset contains multiple versions of piano arrangements of 909 popular songs composed by professional musicians.

The main body of the dataset contains the vocal melody, lead instrument melody, and piano accompaniment of each song in MIDI format, which are aligned with the original audio files. In addition, the dataset also provides annotations of speed, beat, key, and chord, of which the speed curve is manually annotated and the others are completed by the MIR algorithm.