VideoRun2D: Cost-Effective Markerless Motion Capture for Sprint Biomechanics

Abstract: Sprinting is a determinant ability, especially in team sports. The kinematics of the sprint have been studied in the past using different methods specially developed considering human biomechanics and, among those methods, markerless systems stand out as very cost-effective. On the other hand, we have now multiple general methods for pixel and body tracking based on recent machine learning breakthroughs with excellent performance in body tracking, but these excellent trackers do not generally consider realistic human biomechanics. This investigation first adapts two of these general trackers (MoveNet and CoTracker) for realistic biomechanical analysis and then evaluate them in comparison to manual tracking (with key points manually marked using the software Kinovea). Our best resulting markerless body tracker particularly adapted for sprint biomechanics is termed VideoRun2D. The experimental development and assessment of VideoRun2D is reported on forty sprints recorded with a video camera from 5 different subjects, focusing our analysis in 3 key angles in sprint biomechanics: inclination of the trunk, flex extension of the hip and the knee. The CoTracker method showed huge differences compared to the manual labeling approach. However, the angle curves were correctly estimated by the MoveNet method, finding errors between 3.2{\deg} and 5.5{\deg}. In conclusion, our proposed VideoRun2D based on MoveNet core seems to be a helpful tool for evaluating sprint kinematics in some scenarios. On the other hand, the observed precision of this first version of VideoRun2D as a markerless sprint analysis system may not be yet enough for highly demanding applications. Future research lines towards that purpose are also discussed at the end: better tracking post-processing and user- and time-dependent adaptation.

• The paper introduces VideoRun2D, a system adapting MoveNet and CoTracker for cost-effective markerless motion capture in sprint biomechanics, evaluated against manual ground truth.
• Evaluation showed MoveNet achieved viable mean joint angle errors between 3.2-5.5° in real-world sprint conditions, while CoTracker proved unsuitable for precision analysis.
• VideoRun2D offers a promising accessible approach for sprint training and rehabilitation analysis, with the evaluation dataset released to aid future refinement and research.

VideoRun2D: Cost-Effective Markerless Motion Capture for Sprint Biomechanics

The paper "VideoRun2D: Cost-Effective Markerless Motion Capture for Sprint Biomechanics" addresses a significant intersection of computing and sports science by introducing a markerless motion capture system tailored for sprint biomechanics. The authors aim to evaluate and enhance existing markerless body tracking systems for application in sprint kinematics—a critical component in various sports.

In the domain of sports biomechanics, accurate motion capture is imperative for performance optimization and injury prevention. Historically, such capture utilized marker-based systems, which, while precise, are resource-intensive and often impractical outside controlled environments. Recent advancements in computer vision and machine learning have introduced markerless alternatives, which are both cost-effective and easier to implement. However, these systems often fail to address the complexity of human biomechanics accurately, particularly in high-precision contexts such as sprint analysis.

Methodology and Evaluation

The study presented in the paper endeavors to bridge the gap between general-purpose markerless tracking systems and the demanding specifications of sprint biomechanics. This is executed through the adaptation and comparison of two advanced tracking systems: MoveNet and CoTracker. The authors introduce VideoRun2D, a system designed to harness these technologies specifically for sprint biomechanics. The evaluation focuses on key locomotive angles: trunk inclination, and hip and knee flexion-extension, as they are central to biomarker assessments in sprint performance.

The experimental setup involves a dataset comprising 40 sprint video recordings from 5 subjects, under real-world conditions with standard video equipment. This setup contrasts manual tracking via the software Kinovea, designated as the ground truth, with the markerless solutions.

Results and Findings

Results demonstrate that MoveNet, at the core of VideoRun2D, aligns closely with ground truth measurements, exhibiting mean errors in joint angle estimation between 3.2° and 5.5°. This level of precision, despite implementation challenges in uncontrolled environments, suggests viability for numerous applications, albeit with limitations for the most exacting analyses.

CoTracker, however, shows substantial deviations from the ground truth, indicating misalignments unsuitable for precision-demanding scenarios in sprint biomechanics assessment. The error margin with CoTracker spans between 16° and 61.8°, dependent on the joint in question, which suggests substantial limitations for its use in applied biomechanics without further calibration or methodological improvements.

Implications and Future Directions

This research provides insightful perspectives on the practical use cases of markerless systems in athletic performance analysis. The positive outcomes with MoveNet are promising, suggesting that such systems can support more accessible and scalable biomechanical analysis useful for sports professionals and biomechanics researchers.

However, to reach the precision of traditional marker-based systems, future iterations should manage variability across individuals and time-specific assessment errors more effectively. Enhancing post-processing techniques, and exploring adaptive methodologies to improve accuracy dynamically, are identified as vital developmental paths.

Furthermore, the release of the dataset used in this evaluation holds significant value, supporting the community with granular biomechanical data to refine and innovate within markerless capture technologies.

Overall, VideoRun2D positions itself as a viable preliminary model for cost-effective, accessible sprint biomechanical examination, holding particular promise in the contexts of sports training and rehabilitation, where flexibility and lower overheads are advantageous. Continued research and refinement in this field could refine its robustness and applicability, setting a foundation for more comprehensive integration in athletic environments.