Multimodal Golf Swing Motion Data with Sensor Fusion

1,557 golf swing samples from 24 participants of varying skill levels. It integrates synchronized data streams from a full-body motion capture system, 17 IMUs, custom insole pressure sensors, first-person view videos with gaze tracking, and third-person view RGB-D videos. Shot outcome metrics like carry distance and ball speed are provided as metadata.

Use Cases

• Train sensor fusion models for 3D human pose estimation using synchronized data from 17 IMUs and a motion capture system.
• Analyze the correlation between insole pressure sensor data and shot outcome metrics like carry distance to study swing mechanics.
• Develop models for swing phase segmentation using frame-level timestamps across heterogeneous sensor streams.
• Use first-person view videos with gaze tracking to analyze visual attention patterns during the golf swing.

Strengths

• Contains 1,557 swing samples, providing a substantial collection for motion analysis.
• Integrates data from 24 participants ranging from beginners to professionals, offering skill-level diversity.
• Provides synchronized multimodal streams including IMUs, pressure sensors, and multiple video views.
• Includes post-hoc temporal alignment across sensors to support machine learning tasks.

Limitations

• The participant pool of 24 individuals may limit the generalizability of findings to a broader population.
• As a specialized golf motion dataset, its applicability to other human movement domains is untested.
• The dataset's complexity from multiple sensor types may require significant preprocessing and computational resources for fusion tasks.

Provenance

Source

Harvard Dataverse

Collection Method

Collected using synchronized wearable sensors, motion capture, pressure insoles, and video recording during golf swings.

Freshness

Last updated in February 2026.