rPPG-10

This dataset comprises facial video recordings and it was specifically created to support research in remote (non-contact) photoplethysmography (rPPG). It aims to facilitate the development and validation of non-contact heart rate estimation methods from facial videos. The dataset includes data from 26 healthy Portuguese university students (12 male and 14 female), with a mean age of 22.5 ± 1.2 years. Although a total of 27 subjects were originally recorded, one participant (Subject_4) was excluded from the final dataset due to significant artifacts in both video and ECG data.

For each subject, a video recording lasting 10 minutes (600 seconds) was captured, which was then divided into 3 regions of interest (ROI): the forehead, the right cheek (Cheek1), and the left cheek (Cheek2). These videos are provided in AVI format with a resolution of 64×64 pixels and follow a standardized naming convention: Subject_X_ROI.avi (e.g., Subject_1_Forehead.avi). Alongside the video data, a synchronized electrocardiogram (ECG) signal is available for each subject. These signals are stored in NumPy .npy format for easy access using the Numpy library or compatible tools. They also followed a standardized naming convention: Subject_X_ECG.npy (e.g., Subject_1_ECG.avi). Each subject's data is organized into a dedicated folder (e.g., Subject_1, Subject_2, etc.), which contains the three ROI-specific videos and the associated ECG file.

Additionally, an Excel spreadsheet (.xlsx) is included, containing metadata for each participant. This file documents demographic details (such as age, sex, height, weight, and Fitzpatrick skin type), environmental conditions during the recording (including temperature, humidity, and illumination), as other variables that could influence heart rate or signal quality. These include physical activity habits, recent intake of alcohol, smoking status, and the presence of facial obstructions (e.g., makeup or sunscreen).

This dataset was recorded under natural lighting conditions, introducing realistic variability that reflects challenging scenarios for rPPG. The presence of synchronized ECG provides a reliable ground truth, making this resource valuable for benchmarking and advancing heart rate estimation algorithms in non-controlled environments.