VicarPPG-2 dataset
2020 · 24.2 GB · license agreement required
60 fps facial video with synchronised ECG and PPG ground truth, recorded for evaluating remote PPG.
Request accessVicarPPG-2 is a video and physiological signal dataset, recorded specifically to evaluate remote PPG (rPPG) algorithms at estimating heart rate and short-term heart rate variability. Short-term HRV needs at least 5 minutes of observation, so every recording is 5 minutes long.
Ten participants sat about 1 m in front of a computer screen showing the instructions, with a webcam on top of the screen and an LED ring lamp behind it. Screen brightness was turned down as far as possible to limit its influence on the face. The room was lit by natural ambient light plus the ring lamp (Falcon Eyes DVR-300DVC), to prevent strong shadows and large changes in light. Ground truth was recorded as synchronised ECG at 250 Hz from an Arduino-based ECG board (AD8232) and synchronised PPG at 60 Hz from a pulse oximeter (CMS50E) on the left index finger.
Each participant was recorded in four conditions:
- Baseline (
basein the file names): sitting naturally while watching a relaxing video. - Stress (
hrv): playing a difficult game based on the Stroop task. - Movement (
mov): head and eye movements at set times, listed below. - Post-workout (
run): cooling down after a fatiguing workout, to reach higher heart rates.
| Time (s) | Movement |
|---|---|
| 0–30 | none |
| 30–60 | tilting the head |
| 60–90 | nodding |
| 90–120 | shaking the head |
| 120–150 | tilting, nodding and shaking |
| 150–210 | natural |
| 210–270 | looking away |
| 270–300 | moving only the eyes |
In total, 200 minutes of video were collected. All participants gave informed consent.
The ground truth is in GroundTruth/ECG and GroundTruth/PPG, each in two versions: Raw (automatically detected peaks) and Cleaned (peaks cleaned by hand, and synchronised with the video by hand). The files are CSV files with the columns Time (ms), Signal and Peaks, named after their video (e.g. 01-base.mp4 and 01-base ECG.csv). Two recordings had a signal too poor to clean, because of finger movement in the PPG device and a detaching ECG electrode: 06-base ECG and 09-hrv PPG are only included raw.
At a glance
| Subjects | 10 (male:female 7:3) |
|---|---|
| Age | 29 ± 5 years |
| Skin types | Fitzpatrick II–IV |
| Video | 40 recordings, 60 fps, Logitech Brio, H.264 in MP4 |
| ECG ground truth | 250 Hz (AD8232) |
| PPG ground truth | 60 Hz, finger pulse oximeter (CMS50E) |
| Conditions | Baseline, movement, stress, post-workout; 5 min each |
| Total | 200 min of video |
| Images in publications | Only of subjects 01 and 04 |
How to cite
Please cite this paper in any work that uses VicarPPG-2:
Gudi A, Bittner M, van Gemert J. Real-Time Webcam Heart-Rate and Variability Estimation with Clean Ground Truth for Evaluation. Applied Sciences. 2020; 10(23):8630. MDPI · arXiv
BibTeX
@article{gudi2020real,
title={Real-Time Webcam Heart-Rate and Variability Estimation with Clean Ground Truth for Evaluation},
author={Gudi, Amogh and Bittner, Marian and van Gemert, Jan},
journal={Applied Sciences},
volume={10},
number={23},
pages={8630},
year={2020},
publisher={Multidisciplinary Digital Publishing Institute}
}