This paper is accepted at the 2026 IEEE International Conference on Face and Gestures, we also have a more detailed arXiv version here

Abstract

The Rehab Pile Datasets

We present below the 21 extrinsic regression and 39 classification datasets in our rehabilitation archive.

In order to download our full archive you can simply execute this bash file.

All information in both tables below can be found in this json file.

In order to read the datasets you can use the following utils functions.

Details for the 21 regression datasets.
Dataset	Exercise	Acronym	Samples	Subjects	Folds	Frames	Joints	Dimensions	Label-Range	Article

Details for the 39 classification datasets.
Dataset	Exercise	Acronym	Samples	Subjects	Folds	Frames	Joints	Dimensions	Classes	Article

BibTeX


  @inproceedings{ismail2026deeprehab-pile-fg,
  author = {Ismail-Fawaz, Ali and Devanne, Maxime and Berretti, Stefano and Weber, Jonathan and Forestier, Germain},
  title = {A Standardized Benchmark for Skeleton-Based Rehabilitation Assessment Using Deep Learning},
  booktitle = {IEEE international conference series on Automatic Face and Gesture Recognition (FG)},
  city = {Kyoto},
  country = {Japan},
  pages = {1--9},
  year = {2026}
}