MindMuse Research Code

This repo trains a GAN and then receives input from a Muse 2 headset to control the GAN in real-time.

The headset is a BCI that uses EEG to record brain activity. This activity is then processed and returned as brain wave bands to a generative adversarial network. The brain data drives the GAN's image output.

Please note it is ongoing research. Also note it has been developed and tested only on Windows, with Python 3.9.

Set Up

1. Install requirements

conda create -n mindmuse python=3.9
conda activate mindmuse
pip install -r requirements.txt

2. Train GAN or resume from checkpoint (if required)

Modify train.py and data.py as needed and then run

python train.py

3. Run real-time generation

Modify inference.py to use desired checkpoint

python start_stream.py # (run as separate process)
python run.py