Unobtrusive detection of Parkinson’s disease from multi-modal and in-the-wild sensor data using deep learning techniques

Supplementary code and pre-trained models for the paper published in Scientific Reports.

Dataset can be found here

Instructions

Preparatory steps

git clone https://github.com/alpapado/deep_pd
cd deep_pd
conda env create -f environment.yml
wget https://zenodo.org/record/4311175/files/data.zip
unzip data.zip -d data

Command to reproduce SData experiment (using pre-trained models)

python deep_pd_mil.py with seed=42 train_params.evaluation_on=sdata imu_params.checkpoint=True typing_params.checkpoint=True

Command to reproduce GData experiment (using pre-trained models)

python deep_pd_mil.py with seed=42 train_params.evaluation_on=gdata imu_params.checkpoint=True typing_params.checkpoint=True