Hugging Face's logo

Datasets:
madteam
/
thinking-bciciv2a

Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Vietnamese
Size:
100K - 1M
Tags:
eeg
motor-imagery
bci
bciciv-2a
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
thinking-bciciv2a / README.md
madteam's picture
madteam
Publish pinned BCICIV-2a research shard
1cd0dea verified
preview code
|
Raw History Blame Contribute Delete
2.91 kB
metadata 
license: other
language:
  - en
  - vi
tags:
  - eeg
  - motor-imagery
  - bci
  - bciciv-2a
pretty_name: BCICIV-2a EEG Motor Imagery
size_categories:
  - 100M<n<1G

BCICIV-2a EEG Motor Imagery

Provenance

This repository contains one consolidated CSV shard derived from the local research copy identified as the Kaggle dataset aymanmostafa11/eeg-motor-imagery-bciciv-2a, which describes the BCI Competition IV 2a motor-imagery recordings associated with the Graz BCI laboratory. The uploader has represented that they have the right to redistribute this research copy. The repository owner must preserve any upstream attribution and terms that apply to the source dataset.

This card does not assert a SPDX license. license: other is intentional until the upstream terms are independently mapped to a recognized license identifier.

Contents

  • BCICIV_2a_all_patients.csv
  • 2,448 epochs from 9 subject identities
  • 22 EEG channels per epoch
  • 201 samples per epoch at 250 Hz
  • labels: left, right, foot, tongue
  • identity fields: patient (subject) and epoch (session boundary)

The CSV is a tabular, already-windowed representation. It must not be pooled with Brennan, D01, Chisco, D09, or SparrKULee.

Intended use

Research and reproducibility testing for the THINKING EEG pipeline, including schema validation, checksum verification, leakage-safe subject/session splits, and controlled training experiments.

Limitations and safety

  • This dataset is not a clinical diagnostic resource.
  • The scientific gate for the current BCICIV-2a encoder is false; no deployability or scientific headline is claimed by this repository.
  • Do not infer identity beyond the explicit patient and epoch fields.
  • Do not use random epoch splits for scientific evaluation.
  • Consumers must verify the immutable catalog revision and shard SHA-256 before training.
  • The repository contains EEG-derived human-subject data. Users are responsible for complying with the source terms, applicable law, institutional review requirements, and any restrictions imposed by the data custodian.

Integrity

Measured source shard:

  • bytes: 214848010
  • SHA-256: fa52c442278b63c9add50939b453f0668719ff1137dd596b5b1cac90fe3aa6cd

The GitHub catalog must pin the actual 40-character Hugging Face commit SHA; main and other moving revisions are not valid for scientific runs.

Citation and attribution

Cite the original BCI Competition IV 2a dataset and the upstream Kaggle source when using this derived tabular copy. Do not represent this repository as the original competition distribution.

Release status

gate_pass: false for the current scientific evaluation. This publication provides versioned data access only; it does not certify model quality or clinical/deployment readiness.