Automated Remote Pulse Oximetry System (ARPOS) Dataset

The Automated Remote Pulse Oximetry dataset consists of colour and IR (infrared) images and distance from camera taken from forty (40) participants. The data consists of regions of interest (ROIs) such as the forehead, cheeks and lips. Participant data has been collected from a real-life environment that is their homes, rather than within a lab setting, as the study was conducted during the covid-19 pandemic. The participants downloaded the ARPOS app developed to acquire participant face data for the research study. Data collected presents various factors such as different illumination within a home environment, different skin pigmentations, genders, age groups and participants with beards. One participant (PIS-3252) also has a nude colour lipstick on. The data was collected to test the hypothesis that heart rate and blood oxygen can be measured remotely using a camera-based system. The ground truth was collected using a commercial pulse oximeter device (Wellue). Alternatively, a few participants used the apple watch series 6 to acquire ground truth data. Different FPS has been collected for participants due to various computing hardware and its processing power.

The ground truth vital data (heart rate and blood oxygen) .txt files are aligned with image time frames. ROI images are named as timestamps, i.e., each ROI is acquired for a timestamp. The image name is formatted as 'cropped-' + Hour + '-' + Minutes + '-' + Second + '-' + MilliSecond. For example, cropped-0-42-11-427. The data collected for each participant is in two resting states (Resting1, Resting2) and one active state (After Exercise).