CoolFace
Datasetpublic

braindecode/chbmit

EEG Dataset This dataset was created using braindecode, a deep learning library for EEG/MEG/ECoG signals. Dataset Information Property Value Recordings 683 Type Windowed (from Epochs object) Channels 17 Sampling frequency 200 Hz Total duration 4 days, 6:51:14 Windows/samples 37,046 Size 12.82 MB Format zarr Quick Start from braindecode.datasets import BaseConcatDataset # Load from Hugging Face Hub dataset =… See the full description on the dataset page: https://huggingface.co/datasets/braindecode/chbmit.

sourceHugging Faceunknownupdated 6mo agoView on Hugging Face
0likes383downloads
README.md103 linesDownload Raw Back to root
1---2tags:3- braindecode4- eeg5- neuroscience6- brain-computer-interface7- deep-learning8license: unknown9---10 11# EEG Dataset12 13This dataset was created using [braindecode](https://braindecode.org), a deep14learning library for EEG/MEG/ECoG signals.15 16## Dataset Information17 18| Property | Value |19|----------|------:|20| Recordings | 683 |21| Type | Windowed (from Epochs object) |22| Channels | 17 |23| Sampling frequency | 200 Hz |24| Total duration | 4 days, 6:51:14 |25| Windows/samples | 37,046 |26| Size | 12.82 MB |27| Format | zarr |28 29## Quick Start30 31```python32from braindecode.datasets import BaseConcatDataset33 34# Load from Hugging Face Hub35dataset = BaseConcatDataset.pull_from_hub("username/dataset-name")36 37# Access a sample38X, y, metainfo = dataset[0]39# X: EEG data [n_channels, n_times]40# y: target label41# metainfo: window indices42```43 44## Training with PyTorch45 46```python47from torch.utils.data import DataLoader48 49loader = DataLoader(dataset, batch_size=32, shuffle=True, num_workers=4)50 51for X, y, metainfo in loader:52    # X: [batch_size, n_channels, n_times]53    # y: [batch_size]54    pass  # Your training code55```56 57## BIDS-inspired Structure58 59This dataset uses a **BIDS-inspired** organization. Metadata files follow BIDS60conventions, while data is stored in Zarr format for efficient deep learning.61 62**BIDS-style metadata:**63- `dataset_description.json` - Dataset information64- `participants.tsv` - Subject metadata65- `*_events.tsv` - Trial/window events66- `*_channels.tsv` - Channel information67- `*_eeg.json` - Recording parameters68 69**Data storage:**70- `dataset.zarr/` - Zarr format (optimized for random access)71 72```73sourcedata/braindecode/74├── dataset_description.json75├── participants.tsv76├── dataset.zarr/77└── sub-<label>/78    └── eeg/79        ├── *_events.tsv80        ├── *_channels.tsv81        └── *_eeg.json82```83 84### Accessing Metadata85 86```python87# Participants info88if hasattr(dataset, "participants"):89    print(dataset.participants)90 91# Events for a recording92if hasattr(dataset.datasets[0], "bids_events"):93    print(dataset.datasets[0].bids_events)94 95# Channel info96if hasattr(dataset.datasets[0], "bids_channels"):97    print(dataset.datasets[0].bids_channels)98```99 100---101 102*Created with [braindecode](https://braindecode.org)*103