EEGDash Corpus#

EEGDash Corpus is a training-ready collection of public EEG datasets for pretraining EEG foundation models, served through EEGDash and converted once with one recipe. Its release, EEGDash Corpus v0, holds 77 datasets (14,074 hours, 7,272 subjects): of the 93 datasets REVE’s paper and release card name, 77 are converted and released, 12 cannot be redistributed (8 blocked by their licences, 4 gated by access agreements) and 4 are excluded.

77datasets converted and released (v0)
14,074hours (v0)
7,272subjects (v0)
8blocked (No Derivatives or no licence)
4gated (distribution forbidden)
4excluded
93datasets tracked (REVE's paper and release card)
77 of the 93 datasets tracked are converted and released in v0; 12 cannot be redistributed (8 blocked by the licence, 4 gated by an access agreement) and 4 are excluded.
Introduction

The corpus

  • What the corpus is and why it exists

  • REVE’s corpus as the starting point

  • How it was built, checked and licensed

  • The 93 datasets tracked: why 12 cannot be redistributed and 4 are excluded

What you will find: the definitions every page uses, the provenance of each number, the licences, the table of every dataset tracked with the reason for each one not in the corpus.

Introduction
EEGDash Corpus v0

The release

  • 77 datasets, 14,074 hours, 7,272 subjects

  • 641 electrode positions in 632 channel sets

  • Hours, channels, subjects, rates, paradigms, recordings, ages and sex, tokens, montages

  • How to download and load it

What you will find: the converted Zarr store measured, the electrode map in 2-D and 3-D, the download, the table of its datasets.

EEGDash Corpus v0