In a recent sports science milestone, Colin Bond (Director of Sanford Health’s Center for Biomechanics and Movement Science), and his cross-disciplinary team completed the motion capture feat of recording hundreds of marathon runners under race conditions: outdoors, ready to perform, and maybe hurting just a bit.
Specifically, the Sanford team showed up to this year’s Sioux Falls Marathon. With them: motion capture cameras and 3D pose estimation technology. Then, in a one-of-a-kind study, they went on to record more than 350 marathoners as they reached mile twenty of the race. This is the Sanford Stride Project. All runners receive a biomechanical overview of their race mechanics and those who opt in will also have their data published as the first instalment for a state-of-the-art open data set.
How did it all come together? A talented biomechanics team using hardware and software from three close partners from the motion capture industry.
Part One: Outside the Lab on Race Day
The big day started early for Colin, biomechanist Zach Granatowicz, and exercise physiologist Jason Dorman. They set up and calibrated their eight Qualisys Miqus Video Plus cameras. These cameras feature wide-angle lenses providing higher resolution and larger field of view than traditional lenses, helping ensure a large enough capture volume on the race course. Extensive in-lab testing validated camera placement so that runners could be seen well enough and to simplify calibration.
Still, working outside the lab comes with its own challenges and calibration at sunrise was, in Colin’s own words, tough. Tough but doable as the team handled glare, changing light, and the rising sun to ensure the cameras were ready.
One unexpected challenge: crossing the course with cables to allow cameras to be positioned on both sides of the course. In the end, the team found a low-tech solution. With a tripod placed in the bed of a pick-up truck, data and power lines ran with a safe clearance for runners of every height.
Part Two: Biomechanical Measurement with Markerless Motion Capture
Turning the cameras on for the entire race would have resulted in very unwieldy video file sizes. Instead they collected video footage over 200 separate sessions, each containing multiple runners. With high-quality video footage captured, it was time for pose estimation.
Markerless motion capture systems, like Theia3D, don’t require markers on participants as in traditional marker-based tracking. This speeds up data collection and eliminate errors arising from marker placement. It also opens up new types of data collections and new questions that can be answered. Case in point: hundreds of sweaty marathon runners at mile 20 of a real race.
Theia3D uses a proprietary but independently validated deep learning algorithm for pose estimation. This process goes from raw video files to the 3-dimensional position and orientation of each runner’s body. What comes out is a series of measurements as well as labelled video files which included the popular skeleton overlay.
Part Three: Managing and Analysing Biomechanical Data
The Sanford team collected a total of 359 racers. This is a non-trivial amount of data to manage and analyse, so they used HAS-Motion’s Sift application as the third part of this process. Key features for this part was Sift’s inclusion of the Visual3D computational engine and its batch processing abilities.
Analysis involved defining and calculating key events (like heel strikes), kinematic quantities (like joint angles), and spatiotemporal metrics (like stride length and flight time) to describe how each runner was running at mile 20. Sift’s Visual3D engine applied the same model, scaled uniquely to each runner, as well as the same event and signal definitions across participants. This can be an exacting task, but Sift’s batch processing meant the analysis for all runners took under 10 minutes.
Next Steps
Race day is over and the data has been processed. The project has demonstrated how markerless motion capture can be used in the field and how biomechanics can move outside the lab. It also extends the team’s commitment to large scale biomechanical data. Look no further than their annual ACL injury screening project, which in 2025 alone provided free ACL injury screening to 825 middle and high school athletes across six schools (and in only eight weeks)! And as Colin is clear, it takes a team to make efforts like this real. In addition to Zach and Jason, the Sanford team also had Holly McMahon and Kacey Doyle marketing the event to local runners and Katie Jensen managing the whole research program.
All 359 runners recorded have received a report describing their running at mile 20 of the 2025 Sioux Falls Marathon. These reports present easy-to-understand text summaries and illustrations describing key phases of the running stride alongside quantitative measures. This type of outreach project helps Sanford to connect with its local community in a way few healthcare systems can and to highlight an integrated care model that trains healthy athletes and treats injured athletes.
“Since COVID, there’s been a renewed societal focus on health and recovery. People increasingly want to measure how they move, sleep, and perform in their daily lives. Biomechanics is a part of that.” — Colin Bond
The Sanford Stride Project builds towards Colin’s long-term goal of translating biomechanical insights into smarter training and rehabilitation strategies for athletes. There is also the beginning of an open data set containing the anonymized data of 34 runners who opted in. This real-world, high-fidelity motion data is provided with the aim of lowering barriers to entry, accelerating innovation, and growing the next generation of biomechanists. Just with the size, quality, and nature of the data collected at mile 20, this would represent a contribution on its own. But there might also be the inspiration for others to go out and collect their own data in the wild.