E-GMD: Expanded Groove MIDI Dataset, 444 Hours of Human Drumming on 43 Drum Kits with Aligned MIDI and Audio (Google Magenta)
- Provider
- Google Magenta (Lee Callender, Curtis Hawthorne, Jesse Engel)
- License
- CC-BY-4.0 (source)
- Format
- CSV (in sampled files)
- Stored at
- Google Cloud Storage, gs://magentadata/datasets/e-gmd (us)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Speech & Audio
- Source
- https://arxiv.org/abs/2004.00188
About this dataset
E-GMD is a drum transcription dataset of human drum performances with aligned MIDI and audio, and the first human-performed drum transcription dataset with note velocity annotations. It expands the Groove MIDI Dataset: the GMD sequences, played by drummers on a Roland TD-11 electronic kit, were re-recorded through a Roland TD-17 on 43 drum kits ranging from electronic (808, 909) to acoustic sounds, at 44.1 kHz and 24 bits, aligned to the MIDI within 2 ms. It holds 45,537 sequences (1,059 unique) totaling 444.5 hours: train 35,217 sequences (341.4 h), test 5,289 (50.9 h) and validation 5,031 (52.2 h), keeping the GMD splits, with every kit present in every split. The v1.0.0/ folder contains e-gmd-v1.0.0.zip (MIDI and WAV, about 96 GB), a MIDI-only zip and e-gmd-v1.0.0.csv metadata with drummer ID, session, performance ID, style, tempo, beat type, time signature, duration, split, file paths and kit_name. Used for automatic drum transcription (ADT) and expressive drum modeling. License: CC BY 4.0, made available by Google LLC. Cite Callender, Hawthorne and Engel, Improving Perceptual Quality of Drum Transcription with the Expanded Groove MIDI Dataset, arXiv:2004.00188.
How to access this dataset in Python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('magentadata/datasets/e-gmd/', detail=False)[:10]Lists the files with anonymous access. Runs on any machine, with no cloud account.
Data structure
File types in sampled files: CSV
Files
Example file: datasets/e-gmd/v1.0.0/e-gmd-v1.0.0-midi.zip
Analyze this dataset without downloading it
On Scigantic, E-GMD: Expanded Groove MIDI Dataset, 444 Hours of Human Drumming on 43 Drum Kits with Aligned MIDI and Audio (Google Magenta) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Speech & Audio datasets
- CocoChorales: 240,000 Synthesized Four-Part Chorale Mixtures with Per-Instrument Stems, MIDI and Note Expression (Google Magenta)
- MInDS-14: multilingual spoken intent detection
- Nemotron Content Safety Audio Dataset
- Voxpopuli
- Groove MIDI Dataset (GMD): 13.6 Hours of Expressive Human Drumming as MIDI with Synthesized Audio (Google Magenta)
- Earnings-21: entity-dense earnings call ASR benchmark
- FLORAS: long-form speech recognition, translation and summarization benchmark
- EdAcc: Edinburgh International Accents of English Corpus