embneural/sota-of-p300-detection-in-bci-competitions-ii-iii

Cascade xDAWN EEGNet for ERP detection

Cascade xDAWN EEGNet for ERP Detection

Getting Started

Prepare Datasets (BCI Competition II and III)

1. Download Data set IIb: ‹P300 speller paradigm› and Data set II: ‹P300 speller paradigm›. Dataset IIb only contains one subject, while dataset II has two subjects (A and B).
2. Download the true label of tests sets of dataset IIa and the true label of tests sets of dataset II ( A and B). I have downloaded those labels of test sets and put them in the folder.
3. Put those true labels into the root directory of the downloaded dataset, respectively.

Requriments

• pytorch
• mne
• pyriemann
• sklearn
• visdom

Usage

1. Change src_path and root in main.py
2. Run visdom in the console.

   visdom

3. Run main.py in the console.

   python main.py

The softmax with temperature $t$ was used to get the probability of the model's output logits in this project. The larger $t$ is, the lower the confidence level of the model and the smoother the output. $$p_{i} = \frac{exp(z_{i}/t)}{\sum_{j}{exp(z_{j}/t)}}$$

License

Distributed under the MIT License. See LICENSE.txt for more information.