MultiSenseGolf: A Multimodal Wearable Sensor Dataset for Human Pose Estimation and Motion Analysis in Golf Swing

MultiSenseGolf is a multimodal wearable sensor-based golf swing motion dataset designed to support sensor fusion-based 3D human pose estimation (HPE) and related motion analysis tasks. It consists of 1,557 swing samples from 24 participants ranging from beginners to professionals. The dataset integrates synchronized streams from a full-body motion capture system with 17 IMUs, custom insole pressure sensors, first-person view (FPV) videos with gaze tracking, and external third-person view RGB-D videos. Additionally, shot outcome metrics recorded by an optical launch monitor, such as carry distance and ball speed, are provided as metadata. Post-hoc temporal alignment across heterogeneous sensors is established using an event-based time synchronization procedure. The data is released as swing-level segments with frame-level timestamps to support direct use in machine learning tasks.