This is official repo for NEST-rPPG:
The first large-scale domain generalization beachmark for rPPG task.
Very easy to run and enjoy it.
Let's download these data sets: VIPL, V4V, BUAA, UBFC, PURE.
All well-precessed STMap and Labels for your convenience.
We put all the preprocessing in the STMap folder. Examples of VIPL and BUAA are given. See VIPL for data sets with unstable frame rates and BUAA data sets for data sets with stable sampling rates. For details, refer to the Readme.txt file in the folder. All well-precessed STMap and Labels for your convenience.
We provide toy datasets (./NEST-rPPG/STMap.zip) that can be run directly with a single command:
cd ./NEST-rPPG
python train.py -g $gpu_ids$ -t 'VIPL'
All well-processed STMap and groud truth label for your convenience.
We use the heart rate estimator of network to evalate the VIPL and V4V datasets. We use the BVP signal estimator of network to evalate the BUAA, PURE, and UBFC-rPPG.
For VIPL and V4V:
python Eval.py
For BUAA, PURE, and UBFC:
python dataSort.py # save the same video clip to one mat file
cd ./Eval_from_BVP
run Test.m # use matlab to caculate the HR, HRV(LF,HF), and RF for evaluation
rPPG; remote heart rate measurement; domain generalization; domain-label-free approach; NEuron STructure modeling (NEST);agnostic domain generalization.
@InProceedings{Lu_2023_CVPR,
author = {Lu, Hao and Yu, Zitong and Niu, Xuesong and Chen, Ying-Cong},
title = {Neuron Structure Modeling for Generalizable Remote Physiological Measurement},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
pages = {18589-18599}
}
@inproceedings{Sun_2023_DOHA,
author = {Sun, Weiyu and Zhang, Xinyu and Lu, Hao and Chen, Ying and Ge, Yun and Huang, Xiaolin and Yuan, Jie and Chen, Yingcong},
title = {Resolve Domain Conflicts for Generalizable Remote Physiological Measurement},
booktitle = {Proceedings of the 31st ACM International Conference on Multimedia},
year = {2023},
pages = {8214–8224}
}
E-mail: [email protected]
Copyright © 2023, Hao LU.