EEGDash — open catalog of EEG / MEG datasets

The open catalog of EEG / MEG datasets — indexed, described, and loadable with one line of Python.

Welcome to the official Hugging Face org for EEGDash. Raw EEG/MEG recordings are never rehosted here — each dataset on this page is a pointer to its canonical source (OpenNeuro, NEMAR, or the lab that collected it), and EEGDashDataset handles download, caching, and conversion to a PyTorch-ready braindecode object. One CSV drives the whole catalog; every card you see here regenerates from it automatically.

🗺️ Browse the interactive catalog 📚 Docs 💻 GitHub 📦 PyPI

Catalog shape

In numbers: the archive indexes 736 EEG / MEG datasets totalling 40,361 subjects, 222,750 recordings, and 85,298 hours of signal. 600+ are already mirrored on 🤗 and growing daily, sourced from OpenNeuro (546) and NEMAR (190). By recording type: 571 EEG · 73 iEEG · 55 MEG · 22 fNIRS, plus a handful of multimodal combos.

Featured datasets

A handful of representative entries, grouped by population. Every slug links to its HF card; every card links back to the canonical source.

🟢 Healthy / neurotypical	🟠 Clinical populations	🟡 Developmental (HBN)
ds002718 · Visual, 18 subj Face processing (Wakeman & Henson) HF · `Wakeman2015`	ds003800 · Resting, PD EEG in Parkinson’s disease HF	EEG2025r1 · 10 paradigms, 136 subj Healthy Brain Network release 1 HF · `HBN_r1_bdf`
ds000117 · Visual, MEG + EEG Multimodal face processing HF · `WakemanHenson_MEEG`	ds002799 · Clinical monitoring Patient-day recording, dementia HF	EEG2025r10 · 8 paradigms, 533 subj HBN release 10 — 32 GB HF
ds000246 · Auditory, MEG CTF 275-channel MEG HF	ds004551 · iEEG Intracranial recordings, surgical HF	EEG2025r10mini · 20 subj HBN mini release for tutorials HF
ds003061 · Auditory Speech / naturalistic listening HF	ds004598 · Motor Motor paradigm study HF	… 22 HBN releases total browse all HBN

Browse all 736 datasets →

Get started in 30 seconds

pip install eegdash

from eegdash import EEGDashDataset

# Load any dataset in the catalog by its ID…
ds = EEGDashDataset(dataset="ds002718", cache_dir="./cache")

# …or by canonical alias — every known name is a registered class:
from eegdash.dataset import Wakeman2015
ds = Wakeman2015(cache_dir="./cache")

# …or pull a Hub-mirrored, pre-windowed Zarr copy:
from braindecode.datasets import BaseConcatDataset
ds = BaseConcatDataset.pull_from_hub("EEGDash/ds002718")

# EEGDash datasets ARE braindecode datasets — plug into PyTorch unchanged.
from torch.utils.data import DataLoader
loader = DataLoader(ds, batch_size=32, shuffle=True)

Contribute

Missing a dataset? Wrong metadata? The whole catalog regenerates from one CSV — fix once, propagate everywhere. Open an issue or see CONTRIBUTING.md.

Cite

If you use EEGDash in your research, please cite the software entry below (and the companion paper once it’s available):

@software{eegdash,
  title     = {EEG-DaSh: an open data, tool, and compute resource for machine learning on neuroelectromagnetic data},
  author    = {Aristimunha, Bruno and Dotan, Aviv and Guetschel, Pierre and Truong, Dung
               and Kokate, Kuntal and Jaiswal, Aman and Majumdar, Amitrava
               and Shirazi, Seyed Yahya and Shriki, Oren and Delorme, Arnaud},
  year      = {2026},
  version   = {0.6.0},
  license   = {BSD-3-Clause},
  url       = {https://eegdash.org},
  howpublished = {\url{https://github.com/eegdash/EEGDash}}
}

When you use a specific dataset, always follow its upstream citation policy — the link lives in every dataset’s HF card under How to cite.