# Evaluating remote photoplethysmography: A 10-minute video dataset in uncontrolled lighting

## TL;DR

This paper introduces a new dataset of 10-minute facial videos paired with heart rate data, aiming to improve remote heart rate monitoring using webcams in real-world conditions.

## Contribution

The paper provides a long-duration, realistic dataset with ECG references for benchmarking and improving rPPG algorithms.

## Key findings

- A dataset of 10-minute webcam videos and ECG recordings from 26 participants was created.
- Baseline heart rate estimations using standard rPPG methods are included for comparison.
- Metadata like temperature and lighting are provided to support realistic algorithm testing.

## Abstract

Remote photoplethysmography (rPPG) is a technique that enables the extraction of physiological parameters, such as heart rate, from video recordings in a completely non-contact manner. Although widely studied, rPPG research has been hindered by the scarcity of long-duration and complex video datasets recorded in realistic, everyday scenarios. In this work, we present a dataset comprising 10-minute facial video recordings of 26 participants, along with recorded electrocardiograms (ECG) used as a reference signal. The recordings were acquired using a low-cost RGB webcam and a research-grade ECG monitor, under natural daylight and uncontrolled ambient lighting conditions. Videos were manually synchronized with the ECG and cropped to a 64 × 64 pixel resolution to ensure subject anonymization. While participants were predominantly young Portuguese university students, and Fitzpatrick skin types I–IV are represented, this limits generalizability to broader populations. Each recording includes auxiliary metadata such as temperature, humidity, and illumination levels measured at the beginning and end of the session. The dataset aims to support the development and benchmarking of rPPG algorithms under realistic webcam conditions. Baseline heart rate estimations using standard rPPG methods are also provided.