Study at a glance
Summary
Disclaimer: The following summary details independent academic research. Theia3D is intended for research, sports performance and educational use only. It is not cleared as a medical device and is not intended to be used for the diagnosis, treatment, mitigation or prevention of any disease, injury or medical condition.
Editor's note: Theia3D version 2024.1.2, as stated in the paper.
Study Overview
Design
A concurrent validity study comparing full-body segment orientations, joint angles, and center of mass position between marker-based and markerless (Theia3D) motion capture during three two-person ("dyadic") sports-relevant movement tasks, with results benchmarked against equivalent single-person versions of the same tasks.
Participants
Twenty-four participants (19 women, 5 men; age 23.0 years; height 176.6 cm; mass 75.16 kg), organized into 12 pairs, with no movement restrictions reported.
Data Collection
Each pair performed three movement tasks, repeated six times each: a "walk high-five" task at walking pace with low occlusion, where partners crossed paths and exchanged a high-five; a "dribble simulation" task at jogging pace, modeling a soccer attacker dribbling a ball past a defender, introducing occlusion from both a second person and a ball; and a "double sidestep" task at running pace, involving an explosive change of direction as partners passed each other. All participants also performed single-person versions of equivalent movements for comparison. Markerless data were captured using eight Sony RX0-II cameras (60 Hz) processed through Theia3D, and marker-based data were captured simultaneously using sixteen Vicon Vantage V5 cameras (200 Hz) with 39 reflective markers, processed using the Vicon Plug-in-Gait full-body model.
Analysis
Segment orientations, joint angles, and center of mass position were compared between the two systems using measures of both similarity (standardized typical estimate of the error and correlation coefficient) and absolute agreement (root mean square difference and Bland-Altman analysis), with results classified as close-to-perfect, acceptable, or unacceptable against established reference thresholds. Results from the two-person tasks were compared against the single-person task results to assess whether measuring multiple individuals simultaneously affected agreement between the systems.
Key Findings
- Segment orientations showed acceptable or close-to-perfect agreement between markerless and marker-based motion capture across all three two-person tasks, with the hand segment as the main exception, which consistently showed unacceptable agreement.
- Flexion-extension joint angles showed acceptable similarity between the two systems across most joints and tasks, again excluding the wrist, which showed unacceptable similarity and agreement throughout.
- Internal and external rotation joint angles consistently showed the weakest agreement between the two systems across nearly all joints and all three tasks, a pattern the researchers were able to partly trace to the hand and lower-arm segments specifically, rather than attributing it only to differences between the two systems' underlying kinematic models.
- Center of mass position showed close-to-perfect similarity and agreement between the two systems across all three two-person tasks, with only a small, consistent offset in the height dimension.
- When compared to equivalent single-person movement tasks, the two-person tasks showed no meaningful difference in agreement between the markerless and marker-based systems, indicating that occlusion from a second athlete, and in one task a ball, did not measurably reduce the markerless system's measurement accuracy relative to solo movement.
What This Means for Movement Science Researchers and Multi-Athlete and Team Interaction Analysis Teams
For researchers and performance staff interested in studying how athletes move and interact with each other, rather than only in isolation, this study offers an encouraging, practically important finding: markerless motion capture's accuracy held up just as well when measuring two people simultaneously, including during tasks involving occlusion from another person and a ball, as it did for solo movement. This supports using markerless capture for the kind of multi-athlete, interaction-focused biomechanical analysis that marker-based systems are poorly suited to, given their setup demands and vulnerability to marker occlusion in close-contact or multi-person scenarios. The study's specific, joint-by-joint findings also offer a practical roadmap: center of mass and most segment orientations and flexion-extension joint angles can be interpreted with confidence in multi-person team sport contexts, while wrist measures and internal/external rotation angles across most joints should be treated with more caution regardless of whether one or multiple athletes are being measured.
The Role of Theia3D in Movement Analysis
Theia3D was the markerless motion-capture system directly evaluated against marker-based motion capture in this study, specifically for its ability to maintain measurement accuracy when capturing two interacting athletes simultaneously, a scenario representative of real team sport play that most prior markerless validation research had not specifically tested. The study's novel comparison between single-person and two-person task performance offers a useful, practically reassuring data point for applying markerless capture to multi-athlete research and performance contexts, beyond the single-athlete applications markerless systems have most commonly been validated for to date.
Study Limitations and Future Directions
The authors note that because both systems recorded simultaneously, physical markers were visible on the video footage used by the markerless system, which could have either helped or hindered the markerless system's ability to detect anatomical landmarks at those locations; they recommend future studies include repeated measurements with and without markers present to isolate this effect, though this fell outside this study's own aims. They note that there is currently no field-wide consensus on what constitutes an acceptable level of agreement between motion-capture systems in team sports specifically, so the thresholds used in this study, while informed by existing literature, involve some degree of judgment. The study compared two motion-capture systems against each other, neither of which represents a true gold-standard reference, so some differences observed may partly reflect limitations in the marker-based system rather than the markerless system alone; the authors recommend future validation against a method like biplanar fluoroscopic imaging to address this. They also note that averaging results across participants may mask meaningful differences in agreement between specific individuals, pointing to one joint angle that showed a consistent but participant-specific offset between systems, and that the markerless system's lower frame rate relative to the marker-based system could have caused it to miss brief, extreme body positions, potentially overestimating the true difference between the two systems. The authors identify testing with larger groups of athletes in bigger, more match-realistic spaces, and validating markerless motion capture against a gold-standard reference method, as important next steps.
Citation: Oonk, G. A., Kempe, M., Lemmink, K. A. P. M., & Buurke, T. J. W. (2025). Examining the concurrent validity of markerless motion capture in dual-athlete team sports movements. Journal of Sports Sciences. Advance online publication. https://doi.org/10.1080/02640414.2025.2497678
Link: Read the full study
Can Markerless Motion Capture Accurately Measure Two Athletes at Once?
This study of 12 pairs of participants finds that Theia3D markerless motion capture shows acceptable to close-to-perfect agreement with marker-based motion capture for most joint angles and center of mass during two-person sports movements, with no meaningful loss of accuracy compared to the same movements performed solo. If your work involves multi-athlete interaction or team sports performance research, talk to our team about applying Theia3D to your own studies.