moabb.datasets.Peters2026#

class moabb.datasets.Peters2026(subjects=None, sessions=None, *, cap=None, return_all_modalities=False)[source]#

Bases: OpenNeuroMirrorMixin, BaseBIDSDataset

[source]

Dataset Snapshot

Peters2026

BCI-FIT: longitudinal Matrix and RSVP P300 calibration recordings of five participants with ALS using a customized non-implantable communication BCI.

P300 / ERP, 2 classes (Target vs NonTarget)

AuthorsBetts Peters, Daniel Klee, Michelle Kinsella, Yonah Hendin, Tab Memmott, Matthew Lawhead, Srikar Ananthoju, Basak Celik, Scott Spaulding, Barry Oken, Melanie Fried-Oken

🇺🇸 Oregon Health & Science University, US·2026·Betts Peters (petersbe@ohsu.edu)
P300 / ERP Code: Peters2026 5 subjects 6 sessions 23 ch (22 EEG) 300 Hz 2 classes 0.8 s trials

Class Labels: Target, NonTarget

Read Paper Compare Similar Report Issue

Overview

BCI-FIT longitudinal P300 calibration dataset from Peters et al. (2026).

BCI-FIT (BCI Functional Implementation Toolkit) is a customization protocol for non-implantable communication BCIs Five participants with speech and/or physical impairments due to amyotrophic lateral sclerosis (ALS; sub-P1 to sub-P5, demographics withheld by the authors) used the Matrix and RSVP typing layouts of BciPy 2.0 over 6 to 10 visits each (9, 10, 9, 6 and 6). Between visits the system was customized (layout, timing, inquiry length, cap) from typing performance, participant preference, clinical observations and offline simulations.

Task and events. Each calibration inquiry shows the target character (1 s), a cue (0.5 s) and then a stream of characters from a 28-symbol alphabet (A-Z, < backspace, _ space), flashed one at a time; the participant watches for the target. MOABB returns one Target or NonTarget epoch per flash, from the target and nontarget rows of the BIDS events.tsv (each flash is also listed with its character, which :meth:`get_additional_metadata` returns as character). Inquiry-preview and key-press rows are not epoched (the files have no prompt or cue rows). The README describes the default settings (55 inquiries of 14 flashes at 5 Hz, 10% of inquiries without a target, 4 s blank); the released runs follow the per-visit customization:

  • - stimulus onset asynchrony of 83 to 337 ms (about 12 Hz to 3 Hz), most runs at 237 ms (Matrix) or 203 ms (RSVP);
  • 14 flashes per inquiry, or 28 in five Matrix runs of sub-P2 (visits 7 to 10), which have a target in every inquiry;
  • eight runs with an inquiry preview confirmed by a switch press (sub-P2 visits 4 and 5 RSVP, sub-P4 visits 3 to 5), most of them with 46 to 50 inquiries and few or no target-free inquiries.

The 88 calibration runs hold 4,472 inquiries and 66,150 flashes (4,109 Target, 62,041 NonTarget). Seven runs end early and are kept as released: 1 inquiry (sub-P1 visit 8 run 2, sub-P4 visit 5 run 10, sub-P5 visit 5 run 3), 2 or 5 inquiries (sub-P4 visit 5 runs 4 and 1, sub-P3 visit 2 RSVP run 6) and 33 inquiries (sub-P2 visit 7 run 1).

Sessions and runs. One MOABB session per visit, "0" for ses-1 to "9" for ses-10. Runs are named "<k>matrix" or "<k>rsvp", with k the order of the calibration run within the visit (BIDS run order), so analyses can select or split by layout from the run name. 70 runs are Matrix and 18 RSVP; RSVP calibration only occurs in the early visits (sub-P1 1-3, sub-P2 1-6, sub-P3 1-4, sub-P4 1-2, sub-P5 1), later visits are Matrix-only.

Caps and channels. Each visit used one dry-electrode cap (Wearable Sensing): the DSI-24 (19 scalp channels, 10-20 names with T3/T4/T5/T6, plus the A2 ear electrode, typed misc here) or the DSI-Flex (7 channels). sub-P3, sub-P4 and sub-P5 used the DSI-24 throughout; sub-P1 used the DSI-Flex at visit 1 and the DSI-24 afterwards; sub-P2 used the DSI-Flex at visits 1 to 7 and the DSI-24 at visits 8 to 10. The two caps share only Cz, P3, Pz and P4. The DSI-Flex channels are Cz, Oz, P3, P4, PO7, PO8 and Pz, as named in the released files. The README lists FCz instead of Cz, but FCz is the DSI-Flex ground electrode (README and every DSI-Flex *_eeg.json), which cannot also be a recording site; BciPy's own DSI-Flex device definition labels the cap's first sensor Cz, and the BciPy BIDS conversion only keeps channels named in its 10-20 list, which has no FCz. This loader therefore keeps Cz and the standard Cz position. The ERPs and spectra of P1 and P2 could not tell Cz from FCz (ratios to Pz varied more between visits than between the two positions), so the evidence for Cz is documentary.

MOABB evaluations load all subjects and visits into one array, so each variant of this loader returns a single channel set. By default (cap=None) all 40 visits are returned with the four shared channels Cz, P3, Pz and P4. cap="DSI-24" returns the 32 DSI-24 visits (all five subjects; sub-P1 visits 2-9, sub-P2 visits 8-10) with the 19 DSI-24 channels (and A2 with return_all_modalities=True), and cap="DSI-Flex" the 8 DSI-Flex visits (sub-P1 and sub-P2) with its 7 channels. Session keys always number the visit, so cap="DSI-24" starts at session "1" for sub-P1. Positions come from the standard 10-20 template; the release's electrode files hold template positions too, not digitized ones.

Signal. Linked-ear reference ((A1 + A2) / 2) with an active common-mode follower; ground at FCz (DSI-Flex) or Fpz (DSI-24); hardware band 0.003-150 Hz. The released files are sampled at 300 Hz and are not filtered: the sidecars declare no software filter, and the recordings carry power above 20 Hz and, in most runs, 60 Hz line noise. The README's downsampling to 150 Hz, 1-20 Hz band-pass and 60 Hz notch were therefore not applied to the distributed files; MOABB paradigms apply their own band-pass.

Timing and the late RSVP P300. Flash onsets are BciPy's display-flip timestamps, shifted during the BIDS conversion by the static offset of the device (0.1 s), the same for both layouts; the release has no trigger or photodiode channel. In the 16 visits with both layouts, the Pz Target minus NonTarget wave peaks later in RSVP than in Matrix runs (cross-correlation delay 50 to 327 ms, median 92 ms; per participant the peak is at 380-450 ms in Matrix and 456-713 ms in RSVP runs), while the flash-locked steady-state visual response has the same latency in both layouts (median difference 7 ms, 95% CI -19 to 16 ms, in the 13 visits at 203 vs 237 ms SOA). A trigger offset would delay both, so the late RSVP P300 is a genuine latency difference and the events are not shifted. The [0, 0.8] s window still contains it.

Not included. The CopyPhrase runs (copy-spelling) are excluded: they are closed loop, with the classifier choosing the next inquiries and stopping at a 0.8 probability threshold, so they have no fixed Target/NonTarget design. The NEMAR mirror on007720 of the OpenNeuro release is used when available, with OpenNeuro as fallback; only the calibration files are downloaded (about 0.9 GB of the 2.3 GB release).

Please cite the BCI-FIT poster (the authors' requested citation) and the OpenNeuro release Release 1.0.1 only changed text relative to 1.0.0; OpenNeuro has registered a DOI for 1.0.0 only.

:param subjects: Subjects to load (1-5 for sub-P1 to sub-P5). :type subjects: list of int | None :param sessions: Sessions (visits) to keep, "0" being ses-1. :type sessions: list of int | str | None :param cap: None returns all visits with the channels both caps share (Cz, P3, Pz, P4); a cap name returns only that cap's visits with its full channel set. Each variant has its own code (Peters2026, Peters-dsi24-2026, Peters-dsiflex-2026), so caches and results stay apart; downloads are shared. :type cap: None | "DSI-24" | "DSI-Flex" :param return_all_modalities: Also return non-EEG channels (A2 with cap="DSI-24"). :type return_all_modalities: bool | dict

Open in Overview tab →

Citation & Impact

Stimulus Protocol
../_images/Peters2026.svg

0.8s task window per trial · 2-class p300 / erp paradigm · 2 runs/session across 6 sessions

HED Event Tags
HED tags2/2 events annotated

Source: MOABB BIDS HED annotation mapping.

Experimental-stimulus
2
Sensory-event
2
Visual-presentation
2
Non-target
1
Target
1
Target
Sensory-eventExperimental-stimulusVisual-presentationTarget
NonTarget
Sensory-eventExperimental-stimulusVisual-presentationNon-target

HED tree view

Tree · Target
├─ Sensory-event
├─ Experimental-stimulus
├─ Visual-presentation
└─ Target
Tree · NonTarget
├─ Sensory-event
├─ Experimental-stimulus
├─ Visual-presentation
└─ Non-target
Channel Summary
Total channels23
EEG22 (dry)
MISC1
Montagestandard_1020
Sampling300 Hz
Referencelinked ears ((A1 + A2) / 2), active common-mode follower
Filter{'highpass': 0.003, 'lowpass': 150}
Notch / line60 Hz

This diagram is automatically generated from MOABB metadata. Please consult the original publication to confirm the experimental protocol details.

BCI-FIT longitudinal P300 calibration dataset from Peters et al. (2026).

BCI-FIT (BCI Functional Implementation Toolkit) is a customization protocol for non-implantable communication BCIs [1]. Five participants with speech and/or physical impairments due to amyotrophic lateral sclerosis (ALS; sub-P1 to sub-P5, demographics withheld by the authors) used the Matrix and RSVP typing layouts of BciPy 2.0 [2] over 6 to 10 visits each (9, 10, 9, 6 and 6). Between visits the system was customized (layout, timing, inquiry length, cap) from typing performance, participant preference, clinical observations and offline simulations.

Task and events. Each calibration inquiry shows the target character (1 s), a cue (0.5 s) and then a stream of characters from a 28-symbol alphabet (A-Z, < backspace, _ space), flashed one at a time; the participant watches for the target. MOABB returns one Target or NonTarget epoch per flash, from the target and nontarget rows of the BIDS events.tsv (each flash is also listed with its character, which get_additional_metadata() returns as character). Inquiry-preview and key-press rows are not epoched (the files have no prompt or cue rows). The README describes the default settings (55 inquiries of 14 flashes at 5 Hz, 10% of inquiries without a target, 4 s blank); the released runs follow the per-visit customization:

  • stimulus onset asynchrony of 83 to 337 ms (about 12 Hz to 3 Hz), most runs at 237 ms (Matrix) or 203 ms (RSVP);

  • 14 flashes per inquiry, or 28 in five Matrix runs of sub-P2 (visits 7 to 10), which have a target in every inquiry;

  • eight runs with an inquiry preview confirmed by a switch press (sub-P2 visits 4 and 5 RSVP, sub-P4 visits 3 to 5), most of them with 46 to 50 inquiries and few or no target-free inquiries.

The 88 calibration runs hold 4,472 inquiries and 66,150 flashes (4,109 Target, 62,041 NonTarget). Seven runs end early and are kept as released: 1 inquiry (sub-P1 visit 8 run 2, sub-P4 visit 5 run 10, sub-P5 visit 5 run 3), 2 or 5 inquiries (sub-P4 visit 5 runs 4 and 1, sub-P3 visit 2 RSVP run 6) and 33 inquiries (sub-P2 visit 7 run 1).

Sessions and runs. One MOABB session per visit, "0" for ses-1 to "9" for ses-10. Runs are named "<k>matrix" or "<k>rsvp", with k the order of the calibration run within the visit (BIDS run order), so analyses can select or split by layout from the run name. 70 runs are Matrix and 18 RSVP; RSVP calibration only occurs in the early visits (sub-P1 1-3, sub-P2 1-6, sub-P3 1-4, sub-P4 1-2, sub-P5 1), later visits are Matrix-only.

Caps and channels. Each visit used one dry-electrode cap (Wearable Sensing): the DSI-24 (19 scalp channels, 10-20 names with T3/T4/T5/T6, plus the A2 ear electrode, typed misc here) or the DSI-Flex (7 channels). sub-P3, sub-P4 and sub-P5 used the DSI-24 throughout; sub-P1 used the DSI-Flex at visit 1 and the DSI-24 afterwards; sub-P2 used the DSI-Flex at visits 1 to 7 and the DSI-24 at visits 8 to 10. The two caps share only Cz, P3, Pz and P4. The DSI-Flex channels are Cz, Oz, P3, P4, PO7, PO8 and Pz, as named in the released files. The README lists FCz instead of Cz, but FCz is the DSI-Flex ground electrode (README and every DSI-Flex *_eeg.json), which cannot also be a recording site; BciPy’s own DSI-Flex device definition labels the cap’s first sensor Cz, and the BciPy BIDS conversion only keeps channels named in its 10-20 list, which has no FCz. This loader therefore keeps Cz and the standard Cz position. The ERPs and spectra of P1 and P2 could not tell Cz from FCz (ratios to Pz varied more between visits than between the two positions), so the evidence for Cz is documentary.

MOABB evaluations load all subjects and visits into one array, so each variant of this loader returns a single channel set. By default (cap=None) all 40 visits are returned with the four shared channels Cz, P3, Pz and P4. cap="DSI-24" returns the 32 DSI-24 visits (all five subjects; sub-P1 visits 2-9, sub-P2 visits 8-10) with the 19 DSI-24 channels (and A2 with return_all_modalities=True), and cap="DSI-Flex" the 8 DSI-Flex visits (sub-P1 and sub-P2) with its 7 channels. Session keys always number the visit, so cap="DSI-24" starts at session "1" for sub-P1. Positions come from the standard 10-20 template; the release’s electrode files hold template positions too, not digitized ones.

Signal. Linked-ear reference ((A1 + A2) / 2) with an active common-mode follower; ground at FCz (DSI-Flex) or Fpz (DSI-24); hardware band 0.003-150 Hz. The released files are sampled at 300 Hz and are not filtered: the sidecars declare no software filter, and the recordings carry power above 20 Hz and, in most runs, 60 Hz line noise. The README’s downsampling to 150 Hz, 1-20 Hz band-pass and 60 Hz notch were therefore not applied to the distributed files; MOABB paradigms apply their own band-pass.

Timing and the late RSVP P300. Flash onsets are BciPy’s display-flip timestamps, shifted during the BIDS conversion by the static offset of the device (0.1 s), the same for both layouts; the release has no trigger or photodiode channel. In the 16 visits with both layouts, the Pz Target minus NonTarget wave peaks later in RSVP than in Matrix runs (cross-correlation delay 50 to 327 ms, median 92 ms; per participant the peak is at 380-450 ms in Matrix and 456-713 ms in RSVP runs), while the flash-locked steady-state visual response has the same latency in both layouts (median difference 7 ms, 95% CI -19 to 16 ms, in the 13 visits at 203 vs 237 ms SOA). A trigger offset would delay both, so the late RSVP P300 is a genuine latency difference and the events are not shifted. The [0, 0.8] s window still contains it.

Not included. The CopyPhrase runs (copy-spelling) are excluded: they are closed loop, with the classifier choosing the next inquiries and stopping at a 0.8 probability threshold, so they have no fixed Target/NonTarget design. The NEMAR mirror on007720 of the OpenNeuro release is used when available, with OpenNeuro as fallback; only the calibration files are downloaded (about 0.9 GB of the 2.3 GB release).

Please cite the BCI-FIT poster [1] (the authors’ requested citation) and the OpenNeuro release [3]. Release 1.0.1 only changed text relative to 1.0.0; OpenNeuro has registered a DOI for 1.0.0 only.

param subjects:

Subjects to load (1-5 for sub-P1 to sub-P5).

type subjects:

list of int | None

param sessions:

Sessions (visits) to keep, "0" being ses-1.

type sessions:

list of int | str | None

param cap:

None returns all visits with the channels both caps share (Cz, P3, Pz, P4); a cap name returns only that cap’s visits with its full channel set. Each variant has its own code (Peters2026, Peters-dsi24-2026, Peters-dsiflex-2026), so caches and results stay apart; downloads are shared.

type cap:

None | “DSI-24” | “DSI-Flex”

param return_all_modalities:

Also return non-EEG channels (A2 with cap="DSI-24").

type return_all_modalities:

bool | dict

References

[1] (1,2)

B. Peters, M. Kinsella, D. Klee, M. Lawhead, T. Memmott, S. Spaulding, B. Oken and M. Fried-Oken, “BCI-FIT: Effects of cBCI customization on performance,” poster, 11th International BCI Meeting, Banff, AB, Canada, June 2-5, 2025.

[2]

T. Memmott, A. Koçanaoğulları, M. Lawhead, D. Klee, S. Dudy, M. Fried-Oken and B. Oken, “BciPy: brain-computer interface software in Python,” Brain-Computer Interfaces, vol. 8, no. 4, pp. 137-153, 2021. DOI: 10.1080/2326263X.2021.1878727

[3]

B. Peters, D. Klee, M. Kinsella, Y. Hendin, T. Memmott, M. Lawhead, S. Ananthoju, B. Celik, S. Spaulding, B. Oken and M. Fried-Oken, “BCI-FIT: A customization protocol for communication brain-computer interface systems,” OpenNeuro ds007720, 2026. DOI: 10.18112/openneuro.ds007720.v1.0.0

from moabb.datasets import Peters2026
dataset = Peters2026()
data = dataset.get_data(subjects=[1])
print(data[1])

Dataset summary

#Subj

5

#Chan

4 (19 or 7 with cap)

#Trials / class

12408 NT / 822 T

Trials length

0.8 s

Freq

300 Hz

#Sessions

6-10

Participants

  • Population: patients

  • Clinical population: amyotrophic lateral sclerosis (ALS) with speech and/or physical impairments

  • BCI experience: mixed

Equipment

  • Amplifier: Wearable Sensing DSI-24 or DSI-Flex (dry electrodes)

  • Electrodes: dry

  • Montage: standard_1020

  • Reference: linked ears ((A1 + A2) / 2), active common-mode follower

Preprocessing

  • Data state: raw

  • Notes: Files are distributed unfiltered at the 300 Hz acquisition rate (60 Hz line noise in most runs); the README’s downsampling to 150 Hz, 1-20 Hz band-pass and 60 Hz notch are not applied to the distributed files.

Data Access

Experimental Protocol

  • Paradigm: p300

  • Task type: calibration

  • Tasks: MatrixCalibration, RSVPCalibration

  • Feedback: none

  • Stimulus: matrix or RSVP speller

Notes

Added in version 1.8.0.

__init__(subjects=None, sessions=None, *, cap=None, return_all_modalities=False)[source]#
property all_subjects#

Full list of subjects available in this dataset (unfiltered).

bids_paths(subject, path=None, force_update=False, update_path=None, verbose=None)[source]#

Calibration .vhdr paths of the selected visits, in visit/run order.

convert_to_bids(path=None, subjects=None, overwrite=False, format='EDF', verbose=None, generate_figures=False)[source]#

Convert the dataset to BIDS format.

Saves the raw EEG data in a BIDS-compliant directory structure. Unlike the caching mechanism (see CacheConfig), the files produced here do not contain a processing-pipeline hash (desc-<hash>) in their names, making the output a clean, shareable BIDS dataset.

Parameters:

  • path (str | Path | None) – Directory under which the BIDS dataset will be written. If None the default MNE data directory is used (same default as the rest of MOABB).

  • subjects (list of int | None) – Subject numbers to convert. If None, all subjects in subject_list are converted.

  • overwrite (bool) – If True, existing BIDS files for a subject are removed before saving. Default is False.

  • format (str) – The file format for the raw EEG data. Supported values are "EDF" (default), "BrainVision", and "EEGLAB".

  • verbose (str | None) – Verbosity level forwarded to MNE/MNE-BIDS.

  • generate_figures (bool) – If True, generate interactive neural signature HTML figures in {bids_root}/derivatives/neural_signatures/. Requires plotly (pip install moabb[interactive]). Default is False.

Returns:

bids_root – Path to the root of the written BIDS dataset.

Return type:

pathlib.Path

Examples

>>> from moabb.datasets import AlexMI
>>> dataset = AlexMI()
>>> bids_root = dataset.convert_to_bids(path="/tmp/bids", subjects=[1])

Notes

Use CacheConfig to configure caching for get_data(). Use moabb.datasets.bids_interface.get_bids_root to get the BIDS root path.

Added in version 1.5.

data_path(subject, path=None, force_update=False, update_path=None, verbose=None)[source]#

Get path to local copy of a subject data.

Parameters:

  • subject (int) – Number of subject to use

  • path (None | str) – Location of where to look for the data storing location. If None, the environment variable or config parameter MNE_DATASETS_(dataset)_PATH is used. If it doesn’t exist, the “~/mne_data” directory is used. If the dataset is not found under the given path, the data will be automatically downloaded to the specified folder.

  • force_update (bool) – Force update of the dataset even if a local copy exists.

  • update_path (bool | None Deprecated) – If True, set the MNE_DATASETS_(dataset)_PATH in mne-python config to the given path. If None, the user is prompted.

  • verbose (bool, str, int, or None) – If not None, override default verbose level (see mne.verbose()).

Returns:

path – Local path to the given data file. This path is contained inside a list of length one, for compatibility.

Return type:

list of str

download(subject_list=None, path=None, force_update=False, update_path=None, accept=False, verbose=None)[source]#

Download the calibration files of the given subjects.

get_additional_metadata(subject, session, run)[source]#

Per-flash metadata: label, presented character, layout, BIDS run, cap.

get_block_repetition(paradigm, subjects, block_list, repetition_list)[source]#

Select data for all provided subjects, blocks and repetitions.

subject -> session -> run -> block -> repetition

See also

get_data

Parameters:

  • subjects (List of int) – List of subject number

  • block_list (List of int) – List of block number

  • repetition_list (List of int) – List of repetition number inside a block

Returns:

data – dict containing the raw data

Return type:

Dict

get_data(subjects=None, cache_config=None, process_pipeline=None, n_jobs=1)[source]#

Return the data corresponding to a list of subjects.

The returned data is a dictionary with the following structure:

data = {"subject_id": {"session_id": {"run_id": run}}}

subjects are on top, then we have sessions, then runs. A sessions is a recording done in a single day, without removing the EEG cap. A session is constitued of at least one run. A run is a single contiguous recording. Some dataset break session in multiple runs.

Processing steps can optionally be applied to the data using the *_pipeline arguments. These pipelines are applied in the following order: raw_pipeline -> epochs_pipeline -> array_pipeline. If a *_pipeline argument is None, the step will be skipped. Therefore, the array_pipeline may either receive a mne.io.Raw or a mne.Epochs object as input depending on whether epochs_pipeline is None or not.

Parameters:

  • subjects (List of int) – List of subject number

  • cache_config (dict | CacheConfig) – Configuration for caching of datasets. See CacheConfig for details.

  • process_pipeline (sklearn.pipeline.Pipeline | None) – Optional processing pipeline to apply to the data. To generate an adequate pipeline, we recommend using moabb.make_process_pipelines(). This pipeline will receive mne.io.BaseRaw objects. The steps names of this pipeline should be elements of StepType. According to their name, the steps should either return a mne.io.BaseRaw, a mne.Epochs, or a numpy.ndarray. This pipeline must be “fixed” because it will not be trained, i.e. no call to fit will be made.

  • n_jobs (int) – Number of jobs to run in parallel over subjects (passed to joblib.Parallel). Default 1 (sequential). Per-subject processing (reading, filtering, resampling, epoching) is independent, so this gives a near-linear speedup for datasets with many subjects.

Returns:

data – dict containing the raw data

Return type:

Dict

property metadata[source]#

Return structured metadata for this dataset.

Returns the DatasetMetadata object from the centralized catalog, or None if metadata is not available for this dataset.

Returns:

The metadata object containing acquisition parameters, participant demographics, experiment details, and documentation. Returns None if no metadata is registered for this dataset.

Return type:

DatasetMetadata | None

Examples

>>> from moabb.datasets import BNCI2014_001
>>> dataset = BNCI2014_001()
>>> dataset.metadata.participants.n_subjects
9
>>> dataset.metadata.acquisition.sampling_rate
250.0
sourcedata_path(subject=None, path=None, force_update=False, verbose=None)[source]#

Get the dataset’s original pre-BIDS distribution from NEMAR.

Where data_path() fetches the original files from the upstream host, this fetches the copy NEMAR mirrors under sourcedata/. The files keep their upstream names, so the two are interchangeable in content – but NEMAR stays reachable when the upstream host is slow, rate-limited, behind a bot gate, or retired.

Parameters:

  • subject (int | str | None) – Restrict the download to one subject, resolved through the deposit’s sourcedata_provenance.json. Deposits enriched before that manifest recorded subjects fall back to the whole tree with a warning. When None the whole tree is fetched.

  • path (None | str) – Base path where MOABB stores datasets.

  • force_update (bool) – Re-fetch even when a local copy is present.

  • verbose (bool, str, int, or None) – If not None, override default verbose level.

Returns:

Local path to the sourcedata directory.

Return type:

str

Raises:

  • ValueError – If the dataset declares no nemar_id.

  • moabb.datasets.download.NemarDownloadError – If the download fails or the deposit publishes no sourcedata/.

upstream_source = 'OpenNeuro'#

Name of the upstream host, used in the fallback warning.