CHILL - Challenging Heartrate and Illumination rPPG

This dataset contains preprocessed facial video frames and corresponding raw heart rate labels intended for research on remote Photoplethysmography (rPPG) and heart rate estimation. The preprocessing was conducted using the rPPG-Toolbox, which is an open-source toolkit for rPPG signal extraction and processing.

The dataset consists of numpy arrays stored as .npy files with the following structure:

Input Files: Formatted as participantid_scenario_input0.npy, containing downsampled facial frames of size 36x36 pixels.

Label Files: Formatted as participantid_scenario_label0.npy, containing raw heart rate labels corresponding to the input frames. The labels are not altered from their original form.

The participant IDs are pseudorandom strings to ensure anonymity. In total, there are 23 participants in the dataset.

Scenarios:

The data is categorized into four scenarios, which simulate different environmental conditions affecting heart rate measurement:

0 - LowHR-Bright: Low heart rate in well-lit conditions.

1 - LowHR-Dark: Low heart rate in low-light conditions.

2 - HighHR-Dark: High heart rate in low-light conditions.

3 - HighHR-Bright: High heart rate in well-lit conditions.

Preprocessing Details:

The preprocessing pipeline involves the following steps:

Face Cropping: Faces are detected and cropped from raw video frames.

Downsampling: Cropped face images are downsampled to a resolution of 36x36 pixels to reduce computational complexity while preserving essential features for rPPG signal extraction. This process also ensures anonymization by significantly reducing identifiable facial details while retaining the necessary information for analysis.

All preprocessing steps were conducted using the rPPG-Toolbox, ensuring consistency and reproducibility.

Request access:

If you would like to request access to these files, please reach out to [e-mail]

You need to satisfy these conditions in order for this request to be accepted:

Individuals wishing to use the data set must hold an academic affiliation. Further to this, they have to download and fill out the End User License Agreement (EULA) and submit it to us. It will be available here soon.

This dataset is intended for research purposes only, specifically for developing and evaluating rPPG and heart rate estimation algorithms under varying lighting and heart rate conditions.