OpenCap team, Stanford University
OpenCap
A motion lab in two iPhones
Open-source software from Stanford that estimates 3D joint angles from synchronised video recorded on two or more iPhones, then uses musculoskeletal simulation to estimate joint loads and muscle activity. Videos upload to a web app and are processed in the cloud.
Why it is interesting
This kind of analysis normally needs a motion capture lab with markers and force plates. In the developers' validation, OpenCap's joint angles were within 4.5 degrees of marker-based capture on average, and a clinician ran a 100-person field study about 25 times faster than a lab-based approach at under 1% of the cost.
Who it suits
University researchers, sports science students and clinicians doing non-profit research who want 3D movement data for squats, jumps, cutting or walking outside a lab.
Before you decide
- Free only for academic or non-profit research and education. A coach, club or clinic using it commercially needs a separate arrangement with the developers' company. [1]
- Accuracy depends on the joint and the movement. An independent study of jumps, hops, squats and cutting found the squat had the lowest errors, averaging 7 degrees and reaching 13.6 degrees for some joint angles. [2]
- The developers' headline accuracy comes from ten healthy adults. It works best when every body segment stays in view of at least two cameras, and the standard setup needs at least two iOS devices. [3]
What the research says
PLOS Computational Biology, 2023
OpenCap: Human movement dynamics from smartphone videos
The developers' validation of OpenCap with two iPhones against marker-based capture and force plates in ten healthy adults: joint angle error averaged 4.5 degrees, ground reaction force error 6.2% of bodyweight; hip, pelvis and lumbar angles are hardest to estimate from video.
doi:10.1371/journal.pcbi.1011462
Journal of Sports Sciences, 2024
Validity and reliability of trunk and lower-limb kinematics during squatting, hopping, jumping and side-stepping using OpenCap markerless motion capture application
Independent Griffith University study of OpenCap against marker-based capture in 19 adults: validity and reliability were joint and task specific, with the squat most accurate (mean error 7.0 degrees, range 2.9 to 13.6).
doi:10.1080/02640414.2024.2415233
Original material
- OpenCap: Sophisticated human biomechanics from smartphone videoStanford Report · announcement · 19 Oct 2023
- Validity and reliability of trunk and lower-limb kinematics during squatting, hopping, jumping and side-stepping using OpenCap markerless motion capture applicationJournal of Sports Sciences · peer-reviewed · 23 Oct 2024
- OpenCap: Human movement dynamics from smartphone videosPLOS Computational Biology · peer-reviewed · 19 Oct 2023
- OpenCapOpenCap (Stanford University and University of Utah) · official page
- opencap-org/opencap-core: Main OpenCap processing pipelineOpenCap on GitHub · documentation
- Sophisticated human biomechanics from smartphone videoStanford on YouTube · the maker's video
Keep exploring from OpenCap
- Catapult Vector 8Performance analysis, Injury risk · Satellite and inertial tracking, indoors and out
- FootbonautTechnique, Performance analysis · A robotic cage that fires passes at you
- ForceDecksPerformance analysis, Injury risk · Lab force plates, carried into the gym
- HOVERAir X1 PROMAXTechnique, Performance analysis · A pocket drone that films you riding
- MLM2PROPerformance analysis, Technique · Launch data and a simulator from behind the ball
- Playermaker 2.0Technique, Performance analysis · Counting touches, left foot and right
- AceComputer vision · A robot arm that beats elite table tennis players
- ANYmal badminton robotComputer vision · A four-legged robot that plays badminton
- BotBoxerComputer vision · A punching bag that dodges back
- HOVERAir X1 PROMAXComputer vision · A pocket drone that films you riding
- Veo Cam 3Computer vision · Nobody has to stand behind it
- "Epidermal VR" haptic skinMulti-sport · Soft skin that plays back touch
- AlterG NEO Anti-Gravity TreadmillMulti-sport · A treadmill that takes away part of gravity
- cryo:one⁺Multi-sport · A walk-in room held near -85 °C
- EksoNRMulti-sport · A walking robot for the rehab gym
- ForceDecksMulti-sport · Lab force plates, carried into the gym
- Gx Sweat PatchMulti-sport · A sticker that reads your sweat
- OpenCap: Human movement dynamics from smartphone videos ↗Research on OpenCap · PLOS Computational Biology
- Validity and reliability of trunk and lower-limb kinematics during squatting, hopping, jumping and side-stepping using OpenCap markerless motion capture application ↗Research on OpenCap · Journal of Sports Sciences
- Learning coordinated badminton skills for legged manipulators ↗Research on ANYmal badminton robot, which also uses Computer vision
- Outplaying elite table tennis players with an autonomous robot ↗Research on Ace, which also uses Computer vision