LSTS: Periodicity Learning via Long Short-term Temporal Shift for Remote Physiological Measurement

This repository is the official implementation of the paper LSTS: Periodicity Learning via Long Short-term Temporal Shift for Remote Physiological Measurement.

Dependencies

conda create -n rppg python=3.9
conda activate rppg
pip install -r requirements.txt
# optional
pip install notebook

Train & Validation

1. Change Path/to/XXXX/dataset and Path/to/cache/directory to the actual paths in preprocess.py
2. Run python preprocess.py
3. Change Path/to/XXXX/dataset and Path/to/cache/directory to the actual paths in the config files in ./configs/
4. Runpython ./train.py --config ./configs/lsts_xxxx.yaml --split_idx idx where xxxx is the name of the dataset and idx is the split index ranging from 0 to 4.
5. The training logs are managed using Weights & Biases. Visit the website to check the results.

Citation

@article{lsts,
    author={Jiang, Titong and Ma, Yuan and Li, Jiaqi and Dong, Qing and Ji, Xuewu and Liu, Yahui},
    journal={IEEE Transactions on Circuits and Systems for Video Technology}, 
    title={LSTS: Periodicity Learning via Long Short-term Temporal Shift for Remote Physiological Measurement}, 
    year={2025},
    volume={},
    number={},
    pages={1-1},
    doi={10.1109/TCSVT.2025.3538474}
}

Credit

This project is heavily dependent on the following projects. If you find them useful, please give them a star.

rPPG-Toolbox TPS Heartpy NeuroKit2 PhysBench minGPT