DeepGaitLab: Reconciling Vision-Based Motion Tracking with ISB Standards
Eni Halilaj (Carnegie Mellon University)
To interface recently developed computer vision tools with an open-source biomechanical modeling software, which should facilitate the uptake of markerless motion tracking in biomedicine.
Wide access to clinical gait analysis could transform the treatment of mobility limitations, which reduce personal independence and overall quality of life in nearly a third of Americans. Motion tracking with computer vision algorithms offers a simpler, cheaper alternative to costly optical motion tracking systems used in research laboratories and highly specialized clinics.
Current computer vision algorithms (e.g., OpenPose, SMPLify), however, do not meet International Society of Biomechanics (ISB) standards. OpenSim, the most widely used biomechanical modeling software, maintains physiological plausibility and dynamical robustness in its solutions. It is therefore trusted by biomedical scientists and clinicians to output clinically relevant outcomes. DeepGaitLab will reconcile two classes of open-source software—computer vision algorithms and biomechanical modeling—making gait analysis more accessible for biomechanists and clinicians.