ChatPPG.
Join free
Model

MW-PPG-HR uncertainty-aware CNN

code release

PPGECGIMUOpen codeWeights not released
Code Cite
UnconfirmedCatalog paper date: Not reportedAdded to atlas: Oct 19, 2026
CitationsView DOI record (opens in a new tab)GitHub starsView on GitHub (opens in a new tab)

Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.

Signals
PPG, ECG, IMU
License

none declared (no LICENSE file in repo at the recorded commit; all rights reserved by default)

At a glance
Type
Task model
Modalities
PPGECGIMU
Form factor
Not reported
Parameters
Not reported
Architecture
Not reported
Venue
Not reported
Year
Not reported
arXiv
Not reported
DOI
Not reported
Code
Open (opens in a new tab)
Weights
Not released
Licence
none declared (no LICENSE file in repo at the recorded commit; all rights reserved by default)
Caveats
Not reported
Datasets
Papers
Added
Oct 19, 2026
Updated
Oct 19, 2026

Original description

This project addresses two fundamental challenges in wearable PPG heart rate estimation: fairness (reduced disparities across skin tones and sexes) and trustworthiness (accurate uncertainty estimates that indicate unreliable predictions). We introduce the MW-PPG-HR dataset covering all six Fitzpatrick skin types and develop an uncertainty-aware deep learning pipeline that eliminates demographic parity gaps through calibrated prediction rejection.

The MW-PPG-HR (Multi-Wavelength PPG Heart Rate) dataset is available on Figshare. The dataset includes:

Recommendation: Concrete dropout provides the best calibration–accuracy–efficiency trade-off (MA = 0.12, competitive MAE, single-model inference).

Verbatim fromGitHub· Description from the repository README

Model · at a glance
MW-PPG-HRMW-PPG-HR uncertainty-aware CNNDownstream tasks
Modalities
3
Linked datasets
1
Infographic generated from the model record fields.

How it works

How MW-PPG-HR uncertainty-aware CNN works: Input signals → Model → Downstream tasksInput signals: PPG (Photoplethysmogram); ECG (Electrocardiogram); IMU (Accelerometer, 3-axis). Dataset: MW-PPG-HR. Model: MW-PPG-HR uncertainty-aware CNN (Masked reconstruction).Input signalsPPG: PhotoplethysmogramPPGPhotoplethysmogramECG: ElectrocardiogramECGElectrocardiogramIMU: Accelerometer, 3-axisxyzIMUAccelerometer, 3-axisDataset: MW-PPG-HRDatasetMW-PPG-HRModelMASK → RECONSTRUCTMW-PPG-HR uncertainty-aware CNNMasked reconstruction

Datasets linked in the hub; pretraining vs evaluation roles follow the paper.

Inputs
PPG, ECG, IMU
Approach
code release · no weights

Datasets used· 1

Papers· 1

Cite this model

Cite the publication recorded for this model. Check it against the publisher before submitting.

MW-PPG-HR uncertainty-aware CNN. (n.d.).
@misc{mwppghruncertainmwppghr,
  title         = {{MW-PPG-HR uncertainty-aware CNN}}
}

Record updated Oct 19, 2026

History

  1. Added to the ChatPPG research hub