ChatPPG.
Join free
Model

PPI-Net

inter-beat-interval trajectory head for wrist-PPG HR with per-point uncertainty

PPGIMUICLR 2027 submissionOpen codeWeights: not released
Paper Cite
Under reviewCatalog paper date: Not reportedAdded to atlas: Oct 19, 2026Source

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.

Not a model to adopt for accuracy. Board: job-hr-estimation.

Signals
PPG, IMU
Pretraining data
None (supervised; PPG-DaLiA LOSO)
Form factor
Wrist
License
MIT (code); no weights released
Architecture

Backbone shared with scalar control; 128-point PPI trajectory head with per-point uncertainty, inverse-variance-weighted HR aggregation

At a glance
Type
Task model
Modalities
PPGIMU
Form factor
Wrist
Parameters
Not reported
Architecture
Backbone shared with scalar control; 128-point PPI trajectory head with per-point uncertainty, inverse-variance-weighted HR aggregation
Venue
ICLR 2027 submission (OpenReview pMkLClhQKn); code release states IEEE TBME submission
Year
Not reported
arXiv
Not reported
DOI
Not reported
Code
Open
Weights
Not released
Licence
MIT (code); no weights released
Caveats
Not reported
Datasets
Devices
Added
Oct 19, 2026
Updated
Oct 19, 2026
PPI-Net: ChatPPG research summary card

Key results

4.04
MAE (bpm) ↓ · Heart rate estimation
PPG-DaLiA
#2 / 14

How it works

How PPI-Net works: Input signals → Model → Downstream tasksInput signals: PPG (Photoplethysmogram); IMU (Accelerometer, 3-axis). Dataset: PPG-DaLiA. Model: PPI-Net. Downstream tasks: Heart rate.Input signalsPPG: PhotoplethysmogramPPGPhotoplethysmogramIMU: Accelerometer, 3-axisxyzIMUAccelerometer, 3-axisDataset: PPG-DaLiADatasetPPG-DaLiAModelPPI-NetDownstream tasksHeart rateHeart rate

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

Pretraining data
None (supervised; PPG-DaLiA LOSO)
Architecture
Backbone shared with scalar control; 128-point PPI trajectory head with per-point uncertainty, inverse-variance-weighted HR aggregation
Inputs
PPG, IMU
Approach
HR estimation · wrist PPG · PPI · uncertainty · selective prediction · negative result

Devices· 2

Product photograph of a black Empatica E4 wristband, seen from above
Empatica E4
Empatica · Wrist
PPGIMUEDATemp
Product image: Empatica
RespiBAN chest band
BiosignalsPlux / RespiBAN-class · Chest
ECGIMURespTemp

Full photo credits and licences are on each device page. An icon means we have no openly licensed or manufacturer-cleared photo of that exact device yet.

Leaderboard rows· 1

All leaderboards
TaskDatasetMetricValueRank
Heart rate estimationPPG-DaLiAMAE (bpm) ↓4.04#2 / 14

Datasets· 1

Cite

APA

Structured Trajectory Supervision for Reliable Physiological Time-Series Regression (PPI-Net). (2026). ICLR 2027 submission. https://openreview.net/forum?id=pMkLClhQKn

BibTeX

@misc{ppinet2026structured,
  title         = {{Structured Trajectory Supervision for Reliable Physiological Time-Series Regression (PPI-Net)}},
  year          = {2026},
  howpublished  = {ICLR 2027 submission},
  url           = {https://openreview.net/forum?id=pMkLClhQKn}
}

Built from the hub record. Check it against the publisher before submitting.

History

  1. Added to the ChatPPG research hub