Deep learning-based remote photoplethysmography (rPPG): how to extract pulse signal from video using deep learning tools Source code of the master thesis titled "Deep learning-enabled remote monitoring of pulse rate for versatile patients"
A method of data collection has been devised that uses a mobile and remote photoplethysmography (PPG) setup to obtain image data from people monitored with an RGB camera and pulse sensor. Through deep learning, the pulse rate (PR) is determined in a contactless manner using a self-generated dataset and two publicly available datasets. This study shows that the PhysNet model performs well on its training dataset, which is the UBFC-rPPG dataset. However, validating deep learning models with crossdataset in this study indicates a weakness in the generalizability of the models. In summary, the method proposed for remote-photoplethysmography (rPPG) is yet to be suitable for the reliable determination of PR in versatile patients.
Laurens Kreilinger. "Deep learning-enabled remote monitoring of pulse rate for versatile patients". Master Thesis. Trier: Trier University of Applied Sciences and Rosenheim: Rosenheim University of Applied Sciences. 2022