The target of this project is to classify EEG data recordings with a CNN architecture. As the data provided is insufficient for training up to a high accuracy, we focus on methods to increace learning with few datapoints. The methods applied include PCA, transfer-training and for testing the k-cross validation.
The Project was conducted as a final project in the course "Implementing Artificial Neural Networks with Tensorflow" in 2021/22 at the University of Osnabrueck by Fabienne Kock, Lucas Liess-Duquesne and Sascha Mühlinghaus.
For further information please refer to our report.
We will use a dataset published by Nicolas Nieto, Victoria Peterson, Hugo Rufiner, Juan Kamienkowski, Ruben Spies. A detailed description can be found here.
git clone https://github.com/lucasld/inner_speech_decoding.git
Use this command to download the dataset into the project folder:
aws s3 sync --no-sign-request s3://openneuro.org/ds003626 dataset/
or use these instructions.
The provided 'environment.yml' includes all the required packages and libraries.