State of the art on ChatPPG
#1 on CAPTURE-24 Activity recognition
0.860Cohen's κ (higher is better)
- Method
- ActiNet
- Reported in
- ActiNet: an open-source tool for activity intensity classification of wrist-worn accelerometry using self-supervised deep learning · Table 1
- Compared
- 2 results under the same protocol; runner-up Random forest + HMM (accelerometer package) 0.800
- Protocol
- 4-class activity intensity · nested 5-fold group (participant) CV · within-dataset · fine-tuned + HMM smoothing · 30 s windows
Embed the SOTA badge
Markdown
[](https://chatppg.com/research/models/actinet#key-results)HTML
<a href="https://chatppg.com/research/models/actinet#key-results"><img src="https://chatppg.com/research/badges/actinet--har-capture-24-cohen-kappa.svg" alt="SOTA on ChatPPG: Activity recognition · CAPTURE-24" height="20"></a>Results as reported by the authors; not independently reproduced. Awarded 2026-10-20; re-checked on every atlas sync.