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.
- Form factor
- Wrist
- Signals
- PPG, HRV
- License
none declared (no LICENSE file in repo at the recorded commit; all rights reserved by default)
- Type
- Task model
- Form factor
- Wrist
- Parameters
- Not reported
- Architecture
- Not reported
- Venue
- Not reported
- Year
- Not reported
- arXiv
- Not reported
- DOI
- Not reported
- Weights
- Not released
- Licence
- none declared (no LICENSE file in repo at the recorded commit; all rights reserved by default)
- Caveats
- Not reported
- Datasets
- Not reported
- Added
- Oct 19, 2026
- Updated
- Oct 19, 2026
Original description
WatchSleepNet is a novel deep learning model designed to improve sleep staging using wrist-worn wearables. The model integrates ResNet, Temporal Convolutional Networks (TCN), and LSTM with attention to enhance temporal feature extraction. A key innovation is the pretraining strategy, which first trains on large ECG-derived Inter-Beat Interval (IBI) datasets and then fine-tunes on smaller wrist-worn PPG-derived IBI datasets, significantly improving generalization. WatchSleepNet achieves state-of-the-art performance on the DREAMT dataset, surpassing existing wearable sleep staging models. This repository provides the full implementation and benchmarking tools to support reproducible research and future advancements in sleep monitoring.
Verbatim fromGitHub· Description from the repository README
How it works
- Inputs
- PPG, HRV
- Approach
- code release · no weights
Datasets
Cite
APA
WatchSleepNet. (n.d.).
BibTeX
@misc{watchsleepnetwatchsleepnet,
title = {{WatchSleepNet}}
}Built from the hub record. Check it against the publisher before submitting.
History
Added to the ChatPPG research hub