data sets

symbolic

• KernScores

• Nottingham: original; cleaned

• TAVERN

• Aria Database

• music21 corpus

audio

• GTZAN

• medleyDB

• Free Music Archive (≒ 1 TB)

hybrid

• Bach10

• POP909 (in ISMIR 2020)

• WJazzD

• SymbTr

• Slakh (synthesized audio)

• The Meertens Tune Collections

• URMP Dataset

meta

• CAL500

• CAL10K

• Million Song (280 GB)

• Yale–Classical Archives Corpus (MIDI files on Classical Music Archive)