This is the git repo associated with the article Metaplastic-EEG: Continuous training on brain-signals by
Isabelle Aguilar (Electronic mail: iagu0459@sydney.edu.au.), Thomas Bersani--Veroni, Luis Fernando Herbozo Contreras, Armin Nikpour, Damien Querlioz, and Omid Kavehei
School of Biomedical Engineering, The University of Sydney, NSW 2006, Australia.
├── README.md <- This file
│
├── reports
│ └── ArticleFigures.ipynb <- Generated graphics and figures used in the Article
│
├── requirements.txt <- The requirements file for reproducing the analysis environment on linux.
├── requirements_windows.txt <- The requirements file for reproducing the analysis environment on windows.
│
├── src
│ ├── data
│ │ ├── TUHDEdfToPckl.py <- Script to convert TUH EEG .edf to 12 sec long pckl files.
│ │ └── TUHPcklToNpyMultiprocess.py <- Script to convert TUH EEG pckl files to two x and y npy array used for training
│ │
│ ├── models
│ │ └──StreamTUH.py <- script to train a BNN on a stream of seizure data
│ │ └──SyntheticTaskTUH.py <- script to train a BNN on synthetic datasets for seizure detection.
On linux :
git clone [!!!]
cd EEGDETECTIONWITHMETAPLASTICITY
python -m venv venv --prompt="MetaplasticEEG"
./venv/Scripts/activate
pip install -r requirements.txt
On windows:
git clone [!!!]
cd EEGDETECTIONTWITHMETAPLASTICITY
python -m venv venv --prompt="MetaplasticEEG"
./venv/Scripts/activate
pip install -r requirements_windows.txt
First execute TUHEdfToPckl.py while changing the data folder in/out in the script. Then do the same with TUHPcklNpyMultiprocess.py.
Execute the script associated with the experiments you want to run. The hyperparameters can be changed at the top of each file beside the #Hyperparameter comment. Theses script use wandb to track the logs of the training.
Project based on the cookiecutter data science project template. #cookiecutterdatascience