MixNet

Joining Force of Classical and Modern Approaches Toward the Comprehensive Pipeline in Motor Imagery EEG Classification

Python API and the novel algorithm for motor imagery EEG classification named MixNet. The API benefits BCI researchers ranging from beginners to experts. We demonstrate examples of using the API for loading six benchmark datasets, preprocessing, training, and validating SOTA models, including MixNet. In summary, the API allows the researchers to construct the pipeline to benchmark the newly proposed and recently developed SOTA models.

Website: https://mixnetbci.github.io
Documentation: https://mixnetbci.github.io
Source code: https://github.com/Max-Phairot-A/MixNet
Bug reports: https://github.com/Max-Phairot-A/MixNet/issues

Getting started

Dependencies

Python==3.8.10
tensorflow-gpu==2.7.0
tensorflow-addons==0.16.1
scikit-learn>=1.2.2
wget>=3.2
h5py==3.5.0
pandas>=2.0

Create docker container with dependencies

docker pull tensorflow/tensorflow:2.7.0-gpu
docker run -ti --name mixnet_container docker.io/tensorflow/tensorflow:2.7.0-gpu bash
wget https://github.com/Max-Phairot-A/MixNet/blob/main/requirement.txt
pip install -r requirements.txt

Installation:

Using pip

  pip install mixnet-bci

Using the released python wheel

  wget https://github.com/Max-Phairot-A/MixNet/releases/tag/v1.0.0/mixnet_bci-1.0.0-py3-none-any.whl
  pip install mixnet_bci-1.0.0-py3-none-any.whl

Usage

Tutorial: https://mixnetbci.github.io

Citation

To read & cite our paper, please go to our preprint and our paper.

P. Autthasan, R. Chaisaen, H. Phan, M. D. Vos and T. Wilaiprasitporn, “MixNet: Joining Force of Classical and Modern Approaches Toward the Comprehensive Pipeline in Motor Imagery EEG Classification,” in IEEE Internet of Things Journal, vol. 11, no. 17, pp. 28539-28554, 1 Sept.1, 2024, doi: 10.1109/JIOT.2024.3402254.

@ARTICLE{10533256,
  author={Autthasan, Phairot and Chaisaen, Rattanaphon and Phan, Huy and Vos, Maarten De and Wilaiprasitporn, Theerawit},
  journal={IEEE Internet of Things Journal}, 
  title={MixNet: Joining Force of Classical and Modern Approaches Toward the Comprehensive Pipeline in Motor Imagery EEG Classification}, 
  year={2024},
  volume={11},
  number={17},
  pages={28539-28554},
  keywords={Electroencephalography;Task analysis;Feature extraction;Measurement;Internet of Things;Multitasking;Motors;Adaptive gradient blending;brain-computer interface (BCI);deep learning (DL);motor imagery (MI);multitask learning},
  doi={10.1109/JIOT.2024.3402254}}

Source Code

View source on GitHub