Labeled and raw sensor capture sessions from OpenMuscle wearable devices, published for machine learning research on hand and forearm biomechanics. Free to use under CC BY 4.0 (see License).
This is the first public drop of the corpus the OpenMuscle roadmap commits to: a hosted, downloadable set of hand-biomechanics sessions anyone can train against. It will grow as new capture sessions land.
flexgrid/ FlexGrid V3 capture sessions (CSV, 60-sensor pressure matrix)
merged/ Merged multi-session training CSVs produced by the PC pipeline
curated/ Hand-picked good sessions from early capture runs (Nov 29-30 era)
legacy/ Early prototype captures (txt) from pre-FlexGrid devices
Sessions were recorded with the tooling in
OpenMuscle-Software, which
also reads these formats (openmuscle convert, openmuscle train). A formal
schema document is on the roadmap; until then the loader code in
pc/src/openmuscle/data/ is the format reference.
Empty (zero-byte) capture files from aborted sessions were excluded from this repository.
| Folder | Device | Sensors |
|---|---|---|
| flexgrid, merged | FlexGrid V3 (15x4 Velostat pressure matrix, ESP32-S3) | 60 |
| curated, legacy | Earlier OpenMuscle prototypes (Band / FlexGrid V1 era) | varies |
Ground-truth labels, where present, come from the LASK5 labeling wand or other input devices recorded alongside the sensor stream.
git clone https://github.com/Open-Muscle/OpenMuscle-Dataset.gitThe capture files are plain text/CSV; the clone transfers compressed (well under the on-disk size). For training examples, see the OpenMuscle-Software README.
Recorded sessions with an OpenMuscle device? Open an issue here or on the Hub to discuss adding them. Contributions are credited per the org contributing guide.
This dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You may use, share, and adapt it for any purpose, including commercially, with attribution to the OpenMuscle project.