Model Health is built on marker-less technology, eliminating the need for physical markers while still generating biomechanically meaningful and accurate joint kinematics. The technology builds on OpenCap, developed by our founders while at Stanford University, and was extensively validated in peer-reviewed researchs.
Validation of marker-less technology typically include comparisons with marker-based motion capture in a lab, which is the established gold standard for measuring 3D movement.
Accuracy - how closely the system agrees with marker‑based motion capture - should be assessed at three levels of detail, each adding more granularity:
1. Overall agreement: a single accuracy number summarizing how well the systems agree across tasks and joints.
2. Per-task agreement: accuracy broken down per task (e.g., walking vs running).
3. Per-joint agreement: accuracy for a specific task and joint (e.g., knee flexion during a squat).
Beyond reconstructing 3D motion from video, Model Health automates data analysis for a growing library of tasksthat includes walking, running, jumping, hopping, squatting, cutting, and sprinting. For each type of task, asegmentation algorithm identifies key events such as contact and toe-off times.
To answer these questions precisely, Chris Polley, Sport Science Consultant, gathered the peer-reviewed evidence, our own validation alongside 4 independent research groups that tested the technology across return-to-sport tasks, pathological gait, and treadmill running, in different labs and populations.
The full overview can be downloaded here.