Hugging Face's logo

Datasets:
USuCgex0122e
/
demo_dataset

Dataset card Data Studio Files
xet
 Community
Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Job manager crashed while running this job (missing heartbeats).
Error code:   JobManagerCrashedError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

X
array 2D
y
int32
[[0.10030592978000641,8.043313980102539,14.168291091918945,9.537556648254395,-0.6121148467063904,-3.(...TRUNCATED)
2
[[15.872275352478027,13.076796531677246,11.324017524719238,7.935353755950928,2.540727138519287,0.212(...TRUNCATED)
2
[[8.202235221862793,12.142278671264648,14.74178695678711,15.257962226867676,16.741857528686523,22.51(...TRUNCATED)
2
[[15.665055274963379,19.107282638549805,23.676366806030273,24.68453598022461,22.645002365112305,22.1(...TRUNCATED)
2
[[5.838977336883545,-0.3180580139160156,3.5706706047058105,13.1376314163208,17.927120208740234,15.47(...TRUNCATED)
2
[[17.90625762939453,12.432417869567871,13.997883796691895,16.39369773864746,14.555363655090332,8.903(...TRUNCATED)
2
[[54.164398193359375,56.56955337524414,53.29356384277344,42.09715270996094,29.839622497558594,23.977(...TRUNCATED)
2
[[57.54045867919922,54.74715805053711,48.96881866455078,43.63410568237305,39.61923599243164,37.60305(...TRUNCATED)
2
[[-2.6244089603424072,-2.0931951999664307,-3.854464292526245,-6.9269938468933105,-9.42552375793457,-(...TRUNCATED)
2
[[-0.8132097721099854,-1.1803854703903198,-2.8828377723693848,-5.32390832901001,-5.784379005432129,-(...TRUNCATED)
2
End of preview.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

demo_dataset

This repository contains a curated collection of processed EEG benchmark datasets packaged for model training and evaluation.

Each dataset is stored in its own directory and follows a consistent split-based HDF5 layout:

  • train.h5
  • val.h5
  • test.h5
  • croissant.json
  • croissant_manifest.csv
  • metadata_summary.json

Repository structure

Each dataset directory contains:

  • train.h5, val.h5, test.h5
    • HDF5 files containing the packaged split data.
    • In most datasets, X stores EEG tensors and y stores labels.
    • Some blind-test releases may omit y in specific splits.
  • croissant.json
    • Dataset-level Croissant metadata for discoverability and validation.
  • croissant_manifest.csv
    • Split-level manifest describing file names, sample counts, channel counts, and timepoints.
  • metadata_summary.json
    • Compact summary of task type, label vocabulary, channel count, sampling rate, window length, and split sizes.

Data format

For a typical split file:

  • X.shape = (n_samples, n_channels, n_timepoints)
  • y.shape = (n_samples,) when labels are available

The exact sampling rate, window length, and channel count differ by dataset and are documented in:

  • metadata_summary.json
  • croissant.json

Included datasets

This repository currently includes 60 packaged datasets:

  • AD65
  • ADHD
  • Awakening
  • Awakening_balanced
  • BCIC2A
  • BCIC4_1
  • BCI_Speech
  • BETA_SSVEP
  • BenchmarkSSVEP
  • Broderick_CP
  • Broderick_CP_128
  • Broderick_Rev
  • Broderick_Rev_128
  • CHB_MIT
  • CIRE
  • ChineseEEG2_RA_Tone
  • DEAP
  • DUAL_FREQ_SSVEP
  • Depression_rest
  • EAV
  • EEGMAT
  • EEG_Mortality_PD
  • EEG_SVRec
  • ExoEEG_WalkStop
  • FACED_new
  • HBN_EEG
  • HFO
  • HMC
  • ISRUC_S1
  • ISRUC_S2
  • ISRUC_S3
  • LEMON_age
  • LEMON_extraversion
  • LEMON_gender
  • Longitudinal_EEG_Reliability
  • MDD
  • MODMA
  • MonitoringErrP
  • MusicEEG
  • PD31
  • PEARL_Neuro
  • Physionet_MI
  • RestCog
  • SEED
  • SEEDIV
  • SEED_FRA
  • SEED_V
  • SEED_VIG
  • SEED_VII
  • SHU_MI
  • SSVEP
  • Siena_EEG
  • SleepEDF_full
  • TDBRAIN
  • TUAB
  • TUEP
  • TUEV
  • TUSL
  • Things_EEG2
  • Workload

Metadata policy

The packaged metadata intentionally avoids internal mount paths, usernames, and local infrastructure identifiers.

Where possible, provenance in croissant.json points to public references such as:

  • original dataset papers
  • official dataset release pages
  • PhysioNet / OpenNeuro / institutional dataset pages

Notes

  • This repository is a processed benchmark collection, not a raw-data release.
  • Temporal segmentation is dataset-specific and is not forced to be uniform across all datasets.
  • Some datasets in the same family may use different window lengths if their benchmark packaging differs.
  • For exact task definitions, label mappings, and window parameters, use metadata_summary.json.

Recommended usage

  1. Read metadata_summary.json before training on a dataset.
  2. Use croissant_manifest.csv to inspect split sizes and basic tensor dimensions.
  3. Use croissant.json if you need machine-readable dataset metadata or want to run Croissant validation.
Downloads last month
24
Total file size:
121 GB