Datasets:

File size: 2,609 Bytes
b0aabfd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bbb5950
fa469fe
b0aabfd
 
 
 
 
 
 
 
 
 
fa469fe
b0aabfd
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
---
pretty_name: "PROTEUS BCI Bordeaux \u2014 EEG/EMG Foundation Challenge 2026, Track 02"
license: cc-by-4.0
tags:
  - bids
  - neuroscience
  - eegdash
  - electrophysiology
---

# PROTEUS BCI Bordeaux — EEG/EMG Foundation Challenge 2026, Track 02

This repository hosts the **complete scientific files** of NEMAR `nm000290`,
snapshot `v1.0.0`. It includes every subject and recording, the
BIDS sidecars, and the source's `sourcedata/` and `code/` files where present.
All source files were verified against NEMAR's manifest checksums before upload.
Only DataLad control files and source `.gitattributes` files are omitted because
Hugging Face manages repository storage. The original README is preserved as
[SOURCE_README.md](SOURCE_README.md).

EEG dataset from the PROTEUS brain-computer interface (BCI) project in Bordeaux, released for the EEG/EMG Foundation Challenge 2026 (Track 02). It contains recordings from 10 participants across 14 sessions (112 runs) performing three cued mental tasks: kinesthetic motor imagery, mental subtraction and word generation. Runs include resting baselines, calibration runs with sham feedback and online runs with real feedback, using Graz and BrainHero interfaces.

## Load with the BIDS builder

These examples require [datasets PR #9011](https://github.com/huggingface/datasets/pull/9011)
until its BIDS builder is released:

```bash
pip install "datasets[bids] @ git+https://github.com/bruAristimunha/datasets@feat/bids-biosignal"
```

```python
from datasets import load_dataset


dataset = load_dataset("EEGDash/nm000290", split="train")
row = dataset[0]
recording = row["biosignal"]
print(recording)
samples = recording.get_data(start=0, stop=100)
print(row["metadata"])
```

There are 10 subjects and
112 recordings in the complete mirror.
The example may select one subject; all subjects are stored here.
Accessing a recording caches its subject/session/datatype folder and the repository’s JSON/TSV metadata. Other runs in that folder may also be downloaded. MNE then opens the cached recording lazily.


## Source and attribution

- Source: https://nemar.org/dataset/nm000290
- Snapshot manifest: https://data.nemar.org/nm000290/v1.0.0/manifest.json
- Original authors, license and citations: [source README](SOURCE_README.md) and
  [source metadata](nemar_metadata.json).
- Machine-readable checksums: [nemar_manifest.json](nemar_manifest.json).
- Mirror version and omitted control files: [mirror.json](mirror.json).

Selected using NEMAR's 30-day ranking of **viewer data-chunk bytes**. This ranking
does not measure completed archive downloads.