Miscelaneous MIDI Datasets

---

1) LAKH MuseNet Select Piano MIDI Dataset

~1800 Select MIDIs from LAKH MIDI Dataset that were converted to multi-channel Piano MIDIs for efficient and effective Music AI Piano training.

The main feature of this dataset is that each MIDI has approximately equal pitches density on each channel.

Another notable feature that most MIDIs feature Drums track which further improves Music AI Piano learning/training.

---

2) Ultimate 200 Anime Songs Piano Medley

MIDI transcription of Ultimate 200 Anime Songs Piano Medley YouTube video by Animenz

---

3) POP909 Piano Violin MIDI Dataset

Full POP909 MIDI Dataset in Piano Violin arrangement

---

4) Processed POP909 Piano Violin MIDI Dataset

Full POP909 MIDI Dataset in Piano Violin arrangement but processed into MIDI/TMIDIX ms score format and converted into python dictionary for ease of use

---

5) English Karaoke Melodies

~5.8k processed English karaoke melodies

Format: ['word', 'delta_start_time(4096ms/256)', 'duration(4096ms/256)', 'midi_pitch(1-127)']

---

6) POP909 Piano Mono Violin MIDI Dataset

Same as POP909 Piano Violin MIDI Dataset above but with monophonic Violin melody in each MIDI

---

7) chords-melody MIDI Dataset

chords-melody dataset converted from Music XML to MIDIs

---

8) lead-sheets MIDI Dataset mirror archive copy

lead-sheets dataset original mirror archive copy since the Google Drive link at the original repo does not seem to be working anymore

---

9) www.jsbach.net mirror archive copy

www.jsbach.net mirror archive copy

---

10) ASAP Aligned Scores Performances

Pitches aligned score-performance pairs from ASAP MIDI dataset

---

Project Los Angeles

Tegridy Code 2024