OpenCap Monocular beta release — 3D human motion & forces from 1 phone!
Multi-camera OpenCap helped us move out of the lab, but it still required multiple phones and a laptop. We’ve run studies with hundreds of participants in gyms, clinics, and at patient conferences, but we still had to be there to collect the data, which prevented massive, decentralized biomechanics studies. With OpenCap Monocular, participants can now collect 3D biomechanics data on themselves at home in <2 minutes (no laptop, no calibration, no neutral trial…just one iPhone).
Highlights:
– Global 3D kinematics from a static iPhone computed for free on the OpenCap cloud platform
– Dynamics (ground forces, joint moments, muscle forces) from hybrid physics–ML pipelines
– Optimizes outputs from computer vision pose models to compute biomechanically and physically realistic kinematics (e.g., OpenSim models, realistic foot-floor interaction)
– Example session: https://lnkd.in/gjB8eKhB
Validation:
– Rotational accuracy (4.8° vs. markered mocap, <1° different from 2-cam OpenCap)
– Translational accuracy (3.4cm vs. markered mocap)
– Kinetics accurate enough for several downstream tasks relevant to age-related changes in muscle strength (knee moment during sit-to-stand) and osteoarthritis (joint loading during walking)
Get started:
– Follow our best practices to get good results. Recording setup matters, you may need to iterate. Get started, Best practices, & Troubleshooting tips on opencap.ai.
– We are in beta mode, we are still working to make it work reliably for as many activities as possible. E.g., it does not work for jumping, treadmill walking, or small pediatric populations at the moment… we’re working on it. Leave feedback on our forum.
Excited for our first major software release from the University of Utah Movement Bioengineering Lab!
This work builds on great work in computer vision and biomechanics (e.g., OpenSim, OpenCap, and WHAM).
OpenCap Monocular beta release — 3D human motion & forces from 1 phone!
Multi-camera OpenCap helped us move out of the lab, but it still required multiple phones and a laptop. We’ve run studies with hundreds of participants in gyms, clinics, and at patient conferences, but we still had to be there to collect the data, which prevented massive, decentralized biomechanics studies. With OpenCap Monocular, participants can now collect 3D biomechanics data on themselves at home in <2 minutes (no laptop, no calibration, no neutral trial…just one iPhone).
Highlights:
– Global 3D kinematics from a static iPhone computed for free on the OpenCap cloud platform
– Dynamics (ground forces, joint moments, muscle forces) from hybrid physics–ML pipelines
– Optimizes outputs from computer vision pose models to compute biomechanically and physically realistic kinematics (e.g., OpenSim models, realistic foot-floor interaction)
– Example session: https://lnkd.in/gjB8eKhB
Validation:
– Rotational accuracy (4.8° vs. markered mocap, <1° different from 2-cam OpenCap)
– Translational accuracy (3.4cm vs. markered mocap)
– Kinetics accurate enough for several downstream tasks relevant to age-related changes in muscle strength (knee moment during sit-to-stand) and osteoarthritis (joint loading during walking)
Get started:
– Follow our best practices to get good results. Recording setup matters, you may need to iterate. Get started, Best practices, & Troubleshooting tips on opencap.ai.
– We are in beta mode, we are still working to make it work reliably for as many activities as possible. E.g., it does not work for jumping, treadmill walking, or small pediatric populations at the moment… we’re working on it. Leave feedback on our forum.
Excited for our first major software release from the University of Utah Movement Bioengineering Lab!
This work builds on great work in computer vision and biomechanics (e.g., OpenSim, OpenCap, and WHAM).
Preprint: https://lnkd.in/gzz4U3QN
Video abstract: https://lnkd.in/gyWBWZzp
Website & get started: https://opencap.ai
Project page: https://lnkd.in/ghJenTEb
Open-source code: https://lnkd.in/gZPxs-NU