Reperio-rPPG: Reperio-rPPG: Relational Temporal Graph Neural Networks for Periodicity Learning in Remote Physiological Measurement
This repository is the official implementation of the paper [Reperio-rPPG: Relational Temporal Graph Neural Networks for Periodicity Learning in Remote Physiological Measurement].
conda create -n rppg python=3.9
conda activate rppg
pip install -r requirements.txt
gdown https://drive.google.com/uc?id=1TqE3a7VAVLj0d31YKII_PSerdt2aixBP torch-2.0.0+cu118-cp39-cp39-linux_x86_64
gdown https://drive.google.com/uc?id=1S6rjIxDfawnHROX6EBVD4vLIloMVD2Fc torchvision-0.15.1+cu118-cp39-cp39-linux_x86_64
pip install path/to/torch-2.0.0+cu118-cp39-cp39-linux_x86_64
pip install path/to/torchvision-0.15.1+cu118-cp39-cp39-linux_x86_64
- Replace Path/to/XXXX/dataset and Path/to/cache/directory with the correct directories in preprocess.py
- Execute the script by running:
python preprocess.py - Update the configuration files in ./configs/ by setting Path/to/XXXX/dataset and Path/to/cache/directory to the appropriate paths.
- Start training by executing:
python ./train.py --config ./configs/lsts_xxxx.yaml --split_idx idxwherexxxxrefers to the dataset name andidxis the data split index (from 0 to 4). - Training progress and results are tracked via Weights & Biases. Visit the platform to monitor your experiments.