-
Notifications
You must be signed in to change notification settings - Fork 183
Datasets Support
Bru edited this page May 28, 2023
·
40 revisions
The main object of this page is to maintain a list of datasets available in MOABB and those to add.
To add a dataset in MOABB:
- Indicate in this issue the dataset that you want to include,
- We validate the dataset (verifying the license, the type of data, etc) and add it to this section.
- Open a PR to add the dataset to MOABB!
TODO: add number of trial per class
| Name | Link | Paper | #Subj | #Chan | #Classes | #Trials / class | Trials length | Sampling rate | #Sessions | Size | License | Remarks |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Alex MI | Zenodo | PhD | 8 | 16 | 3 (rh, feet, rest) | 20 | 3s | 512Hz | 1 | 133M | ? | |
| BNCI2014001 | BNCI | Frontiers | 10 | 22 | 4 (rh,lh,feet,tongue) | 144 | 4s | 250Hz | 2 | 744 M | Open access | |
| BNCI2014002 | BNCI | BME | 15 | 15 | 2 (rh, feet) | 80 | 5s | 512Hz | 1 | 848M | Open access | |
| BNCI2014004 | BNCI | TNSRE | 10 | 3 | 2 (lh, rh) | 360 | 4.5s | 250Hz | 5 | 456M | Open access | |
| BNCI2015001 | BNCI | TNSRE | 13 | 13 | 2 (rh, feet) | 200 | 5s | 512Hz | 2 | 1.8G | Open access | |
| BNCI2015004 | BNCI | PLoS | 10 | 30 | 5 (rh, feet, navigation, subtraction, word_ass) | 80 | 7s | 256Hz | 2 | 1.7G | Open access | |
| Cho2017 | GigaDb | GigaScience | 53 | 64 | 2 (lh, rh) | 100 | 3s | 512Hz | 1 | 9.4G | CC0 | |
| Lee2019_MI | GigaDB | GigaScience | 55 | 62 | 2 (lh, rh) | 100 | 4s | 1000Hz | 2 | 61G | CC0 | |
| MunichMI | Zenodo | TBME | 10 | 128 | 2 (lh, rh) | 150 | 7s | 500Hz | 1 | 7.3G | ? | |
| Ofner2017 | Zenodo | PLoS | 15 | 61 | 7 (see below) | 60 | 3s | 512Hz | 1 | 13G | ? | Imagined or executed movements |
| PhysionetMI | Physionet | TBME | 109 | 64 | 5 (lh, rh, feet, hands, rest) | 23 | 3s | 160Hz | 1 | 5.5G | ODC-By | Imagined or executed movements |
| Schirrmeister2017 | Gin | HBM | 14 | 128 | 4 (lh, rh, feet, rest) | 120 | 4s | 500Hz | 1 | 23G | ? | |
| Shin2017A | web | TNSRE | 29 | 30 | 2 (lh, rh) | 30 | 10s | 200Hz | 3 | 5.9G | GPL v3 | includes fNIRS recording |
| Shin2017B | web | TNSRE | 29 | 30 | 2 (substraction, rest) | 30 | 10s | 200Hz | 3 | 5.9G | GPL v3 | mental arithmetics, includes fNIRS recording |
| Weibo2014 | web | PLoS | 10 | 60 | 7 (lh, rh, hands, feet, left_hand_right_foot, right_hand_left_foot, rest) | 80 | 4s | 200Hz | 1 | 4.1G | ? | |
| Zhou2016 | Figshare | PLoS | 4 | 14 | 3 (lh, rh, feet) | 160 | 5 | 250Hz | 3 | 177M | ? |
For Ofner2017, classes are : "right_elbow_flexion", "right_elbow_extension", "right_supination", "right_pronation", "right_hand_close", "right_hand_open", "rest"
| Name | Link | Paper | #Subj | #Chan | #Trials / class | Trials length | Sampling rate | #Sessions | Size | License | Remarks |
|---|---|---|---|---|---|---|---|---|---|---|---|
| BNCI2014008 | BNCI | Frontiers | 8 | 8 | 3500 NT / 700 T | 1s | 256Hz | 1 | 165M | Open access | |
| BNCI2014009 | BNCI | JNE | 10 | 16 | 1440 NT / 288 T | 0.8s | 256Hz | 3 | 177M | Open access | |
| BNCI2015003 | BNCI | NL | 10 | 8 | 1500 NT / 300 T | 0.8s | 256Hz | 1 | 53M | Open access | The first two subject have have a very off target/non-target ratio! |
| DemonsP300 | gin | arXiv | 60 | 8 | 935 NT / 50 T | 1s | 500Hz | 1 | 812M | ? | |
| EPFLP300 | web | JNM | 8 | 32 | 2753 NT / 551 T | 1s | 2048Hz | 4 | 4.3G | ? | |
| Lee2019_ERP | ftp | GigaScience | 54 | 62 | 6900 NT / 1380 T | 1s | 1000Hz | 2 | 86G | CC0 | |
| bi2013a | zenodo | HAL | 24 | 16 | 3200 NT / 640 T | 1s | 512Hz | (S1-S7) 8 sessions / (S8-S24) 1 session | 7.2G | ? |
| Name | Link | Paper | #Subj | #Chan | #Classes | #Trials / class | Trials length | Sampling rate | #Sessions | Size | License | Remarks |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Lee2019_SSVEP | ftp | GigaScience | 24 | 16 | 4 (5.45, 6.67, 8.57, 12Hz) | 25 | 1s | 1000Hz | 1 | 60G | CC0 | |
| MAMEM1 | figshare | arXiv | 10 | 256 | 5 (6.66, 7.5, 8.57, 10, 12Hz) | 12-15 | 3s | 250Hz | 1 | 5.5G | ? | |
| MAMEM2 | figshare | arXiv | 10 | 256 | 5 (6.66, 7.5, 8.57, 10, 12Hz) | 20-30 | 3s | 250Hz | 1 | 4.9G | ? | |
| MAMEM3 | figshare | arXiv | 10 | 14 | 4 (6.66, 8.57, 10, 12Hz) | 20-30 | 3s | 128Hz | 1 | 144M | ? | |
| Nakanishi2015 | GH | PLoS | 9 | 8 | 12 (9.25, 9.75, 10.25, ..., 14.75Hz) | 15 | 4.15s | 256Hz | 1 | 132M | ? | |
| SSVEPExo | zenodo | Neurocomputing | 12 | 8 | 4 (13, 17, 21Hz, rest) | 16 | 2s | 256Hz | 1 | 83M | ? | |
| Wang2016 | ftp | TNSRE | 32 | 62 | 40 (8, 8.2, ..., 15.8Hz) | 6 | 5s | 250Hz | 1 | 6.8G | ? |
To complete
| Name of the Dataset | Description | Link to Dataset | Direct Download | Link to Paper | Type of Paradigm | Number of Subjects | Number of Electrodes | Size of the Dataset | License |
|---|---|---|---|---|---|---|---|---|---|
| Single-flicker online SSVEP BCI datset | 4-class SSVEP | Link | Yes | Link | SSVEP | 12 | 32 | 5.8Gb | Creative Common |
Single stimulus location for two inputs: A combined SSVEP-based brain-computer interface - Data Link
| Name of the Dataset | Link to Dataset | Direct Download | Link to Paper | Type of Paradigm | Number of Subjects | Number of Electrodes | Size of the Dataset | License |
|---|---|---|---|---|---|---|---|---|
| RSA EEG | Link | Yes | Link | Event Related Potential | 10 | 128 | 3Gb | Creative Common |
| David Hubner's ERP dataset | Link | Yes | Link | Event Related Potential | 13 | 31 | 3 Gb in total | Creative Common |
| Medicon 2019 | Link | Yes | NA | Event Related Potential | 15 | 8 | NA | Not Available |
- RSA: 72 different class of images presented for RSA with EEG|
- Hubners: Data from the ERP data for the paper "Learning from Label Proportions in Brain-Computer Interfaces". This method won the Graz BCI 2017 best paper award.
- Medicon: The dataset includes data from 15 participants, with 7 sessions each. It represents the complete EEG recordings of a feasibility clinical trial (clinical-trial ID: NCT02445625 — clinicaltrials.gov) that tested a P300-based Brain Computer Interface to train youngsters with Autism Spectrum Disorder to follow social cues (Amaral et. al, 2017; Amaral et al., 2018). A further description of the experimental setup and design can be found here - The competition dataset is divided into two parts: train and test sets. The train set is available with labels (the target object – out of the 8 different possibilities – for each block) for the contest participants to train their models. The test set is available without labels. The challenge is to predict the labels for each block of the test set. Within each session, the train set consists of 20 blocks and the test set consists of 50 blocks.
Dataset of an EEG-based BCI experiment in Virtual Reality and on a Personal Computer
BEnchmark database Towards BCI Application BETA Dataset , Paper
| Name of the Dataset | Link to Dataset | Direct Download | Link to Paper | Type of Paradigm | Number of Subjects | Number of Electrodes | Size of the Dataset | License |
|---|---|---|---|---|---|---|---|---|
| SEED | Link | No | NA | Affective BCI | 15 | 64 | >10 Gb | Not Available |
- SEED: A dataset collection for various purposes using EEG signals.SJTU Emotion EEG Dataset(SEED)"
Motor Imagery Under Distraction - Paper Link
Rapid Serial Visual Presentation (RSVP) - Paper Link
Openlists for ElectrophysiologyData - Link
Another List of EEG datasets - Link
Take a look at the datasets folder for an example in the paradigm you seek to add a dataset in. Have a look at the tutorial.
Other files to be edited -
- docs/source/dataset.rst
- moabb/datasets/init.py
- (if needed) requirements.txt