Expanded Groove MIDI Dataset (E-GMD)

Quick Start

from datasets import load_dataset

# Stream to avoid downloading the entire dataset
ds = load_dataset("schismaudio/e-gmd", streaming=True)

# Or download locally
ds = load_dataset("schismaudio/e-gmd")

Dataset Description

The Expanded Groove MIDI Dataset (E-GMD) contains 444 hours of aligned MIDI and audio from 43 drum kits on a Roland TD-17 electronic drum kit. The dataset comprises 45,537 sequences derived from 1,059 unique performances by 10 drummers.

E-GMD is an extension of the Groove MIDI Dataset (GMD) that re-renders each original performance through 43 different electronic drum kits, providing massive timbral diversity while preserving the original expressive MIDI data. This makes E-GMD particularly valuable for training drum transcription models that need to generalize across different drum sounds.

Dataset Structure

Data Fields

Field	Type	Description
audio	Audio	WAV audio file
filename	string	Original filename
drummer	string	Anonymized drummer ID
session	string	Recording session identifier
bpm	float	Tempo in beats per minute
beat_type	string	"beat" (steady pattern) or "fill" (fill passage)
time_signature	string	Time signature (e.g., 4-4, 6-8)
duration	float	Duration in seconds
midi	bytes	Raw MIDI data

Data Splits

Split	Examples	Description
train	~36,000	Training set
validation	~5,000	Validation set
test	~4,500	Test set

Splits are partitioned by drummer to prevent data leakage — no drummer appears in more than one split.

Class Taxonomy

The dataset uses raw MIDI pitches from the Roland TD-17 electronic drum kit. The 9 canonical drum classes from the original GMD taxonomy apply:

Class	Instrument	MIDI Pitches
0	Kick	36
1	Snare (Head)	38
2	Snare (Rim)	37, 40
3	Closed Hi-Hat	42, 22
4	Open Hi-Hat	46, 26
5	Low Tom	43, 58
6	Mid Tom	47, 45
7	High Tom	50, 48
8	Ride / Crash	49, 51, 55, 57, 52

Usage Examples

Load audio and MIDI

from datasets import load_dataset

ds = load_dataset("schismaudio/e-gmd")

sample = ds["train"][0]
print(f"Drummer: {sample['drummer']}, BPM: {sample['bpm']}")
audio_array = sample["audio"]["array"]
sr = sample["audio"]["sampling_rate"]

Parse MIDI onsets

import io
import pretty_midi

sample = ds["train"][0]
midi = pretty_midi.PrettyMIDI(io.BytesIO(sample["midi"]))
for instrument in midi.instruments:
    for note in instrument.notes:
        print(f"Time: {note.start:.3f}s, Pitch: {note.pitch}, Velocity: {note.velocity}")

Dataset Creation

Source Data

E-GMD builds on the original Groove MIDI Dataset recorded at Google Magenta. The original performances were captured on a Roland TD-11, with MIDI data preserving exact onset times, pitches, and velocities. For E-GMD, each performance was re-rendered through 43 different drum kits on a Roland TD-17, multiplying the timbral diversity of the dataset by 43x.

Annotations

Annotations are the original MIDI data from the electronic drum kit — no manual annotation was needed. The re-rendering process preserved all timing and velocity information from the original performances.

Known Limitations

• Synthesized audio only: All audio is the electronic drum kit's internal sound engine output, not acoustic drums.
• Shared performances: The underlying MIDI performances are shared across all 43 kit renderings. Models should be evaluated on held-out drummers, not just held-out kit renderings.
• Electronic pad dynamics: Velocity response differs from acoustic drums — pad triggers have fixed dynamic curves.
• Large dataset size: At 444 hours, the full dataset requires significant storage. Consider using streaming mode for exploration.

Related Datasets

This dataset is part of the Drum Audio Datasets collection by schismaudio. Related datasets:

• schismaudio/groove-midi-dataset — The original 13.6-hour GMD with a single kit
• schismaudio/stemgmd — 1,224 hours of isolated per-instrument stems from 10 acoustic kits

Citation

@article{callender2020improving,
  title={Improving Perceptual Quality of Drum Transcription with the Expanded Groove MIDI Dataset},
  author={Callender, Lee and Hawthorne, Curtis and Engel, Jesse},
  journal={arXiv preprint arXiv:2004.00188},
  year={2020}
}

License

This dataset is released under the Creative Commons Attribution 4.0 International License (CC-BY 4.0).

You are free to share and adapt this dataset for any purpose, including commercial use, as long as you give appropriate credit to the original authors (Google Magenta).