yalishanda/e-gmd-v1.0.0-midi

Expanded Groove MIDI Dataset (E-GMD v1.0.0) — MIDI-only mirror

Unofficial mirror. This is a convenience mirror of the MIDI-only portion of the Expanded Groove MIDI Dataset (E-GMD), redistributed under its CC BY 4.0 license (see below). It is not affiliated with or endorsed by Google / Magenta. The canonical source is https://g.co/magenta/e-gmd.

E-GMD is a large dataset of human drum performances (~444 hours) recorded on a Roland TD-17 electronic kit, captured as MIDI with expressive timing and velocity. It extends the original Groove MIDI Dataset (GMD) to many more performances and kits.

What's in this mirror

• e-gmd-v1.0.0-midi.zip — the original Magenta MIDI archive, byte-for-byte (45,537 .midi performances organized by drummer/session).
• e-gmd-v1.0.0.csv — the metadata table (also inside the zip), exposed at the top level for quick previewing. Columns: drummer, session, id, style, bpm, beat_type, time_signature, duration, split, midi_filename, audio_filename, kit_name.
• LICENSE — the original CC BY 4.0 license file.

Not included: the audio renderings (the full E-GMD audio is ~100 GB). The audio_filename column therefore references files not present in this mirror. For audio, use the official source.

Splits

The split column carries E-GMD's official train / validation / test partition (shared across drummers). Use it directly for reproducibility.

License

Provided by Google LLC under a Creative Commons Attribution 4.0 International (CC BY 4.0) license — http://creativecommons.org/licenses/by/4.0/. You may share and adapt with attribution.

Citation

Please cite the E-GMD paper and specify the dataset version (v1.0.0):

@misc{callender2020improving,
    title={Improving Perceptual Quality of Drum Transcription with the Expanded Groove MIDI Dataset},
    author={Lee Callender and Curtis Hawthorne and Jesse Engel},
    year={2020},
    eprint={2004.00188},
    archivePrefix={arXiv},
    primaryClass={cs.SD}
}

Provenance

Mirrored for the Dynamics Needed thesis project (drum-velocity / dynamics prediction). A processed, tabular feature version derived from this data is published separately as yalishanda/dynamics-needed-egmd-tabular.