A Synchronized Dual-System Marker-based and Markerless Motion Capture Dataset for Gait and Fundamental Fitness Activities.

Human motion capture technology is of paramount importance in biomechanical analysis. Although marker-based optical motion capture systems are widely regarded as the gold standard for accurately measuring human movement kinematics, vision-based markerless systems offer greater potential in terms of usability and accessibility. However, the precision of markerless systems in complex human movements requires further validation. The primary aim of this dataset is to enable the comparative evaluation of consistency and accuracy between computer vision-based markerless motion capture systems and traditional marker-based systems. This work presents a synchronized dual-system dataset containing motion capture data from 21 healthy male university students, encompassing seven fundamental fitness activities: slow walking, fast walking, running, squats, lunges, high knees, and jumping jacks. The data were collected synchronously using seven 120 Hz OptiTrack infrared cameras and two 30 Hz smartphones. The dataset comprises 393 valid records, featuring OpenSim-compatible marker trajectory data (.trc format) and 2D coordinate data extracted via YOLOv8 (.csv format). Technical validation was performed to ensure temporal synchronization between the two systems through cross-correlation analysis. In addition, regression analysis and correlation testing were used to assess the consistency of the markerless data for trajectory tracking and action recognition. This dataset establishes a resource for developing and evaluating movement analysis and action recognition algorithms based on both optical marker-based motion trajectories and smartphone-derived markerless keypoint data.