OpenWatch is a multimodal wrist-worn sensor dataset for hand gesture recognition. It captures 59 discrete hand gestures using a custom smartwatch equipped with photoplethysmography (PPG), accelerometer, and gyroscope sensors. The dataset was created by pietrobonazzi and was last updated on 2026-05-06.
Use Cases
- Training gesture recognition models based on multimodal sensor data from PPG, accelerometer, and gyroscope.
- Benchmarking algorithms for discrete hand gesture classification based on the 59 defined gesture classes.
- Developing smartwatch applications that interpret user commands based on wrist-worn sensor signals.
- Studying the correlation between physiological signals (PPG) and physical movements (accelerometer/gyroscope) during gestures.
Strengths
- Includes 59 distinct hand gesture classes, providing a broad range of recognition targets.
- Captures multimodal sensor data from three distinct sources: PPG, 3-axis accelerometer, and 3-axis gyroscope.
- Provides demonstration videos for all gesture classes via a YouTube playlist.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Last updated 2026-05-06 10:37:12; freshness should be verified.
Provenance
- Source
- huggingface
- Collection Method
- Captured using a custom smartwatch equipped with PPG, accelerometer, and gyroscope sensors.
- Freshness
- 2026-05-06 10:37:12