An open-source deep learning model for classifying real-world activity intensities in persons with chronic pain using a wrist-worn accelerometer
Annet Doomen, Richard Felius (equal), Harriët Wittink, Ivan Huijnen, Michiel Punt, Rob Smeets
Summary
A rare CC BY 4.0 wrist-ACC activity-intensity model with weights, so unlike ActiNet (Oxford academic licence) it is usable commercially as a baseline or teacher for Clarity's accelerometer branch. Two caveats: it runs at 12.5 Hz on 10-s windows (coarse), and the promised lab dataset is not actually in the release. Cluster: IMU / HAR.
Key facts
- Form Factor
- wrist
- License
- CC BY 4.0
Reported benchmarks
Human activity recognition (IMU)
capture-24 · Macro-F1 ↑
Sedentary/light/MVPA (no sleep), 10-s windows, Capture-24 downsampled 100→12.5 Hz, trained on CP+HP lab + Capture-24 ("All on Capture"), repeated 10-fold subject-level CV; acc 67.2 %.
Updated: Oct 14, 2026
Related Research
- Human activity recognition (IMU) leaderboard
Benchmark results for human activity recognition (imu)
- AnyMo: Geometry-Aware Setup-Agnostic Modeling of Human Motion in the Wild
Treats wearable setup variation as structured body-surface geometry: dense physics-grounded IMU simu
- BayaHAR: Lightweight Bayesian Few-Shot User Adaptation for On-Device Personalized Human Activity Recognition
Turns pretrained HAR classifiers into prototypical networks with prior prototypes that keep zero-sho
- Bio-PM: Bio-Inspired Self-Supervised Learning for Wrist-worn Accelerometer Data
UMass Amherst / Google Health / UMD team argues that wearable accelerometer SSL has been leaving per