Original title: MuyBridge: Mobile Human Center-of-Mass Estimation from Monocular Video via Sparse Fusion
Aidan Bradshaw, Marco Giordano, David Rode, Andreas Habersack +7 · September 2026
Tracking an athlete's center of mass—their balance point—is vital for sports and medicine. Current tools are too bulky to use outside a lab or fail to follow real human body rules. To solve this, researchers built MuyBridge, a system that tracks this balance point in 3D using standard video from a single phone. MuyBridge works by pairing a small 2D pose model with a fast depth-estimating model. It then combines their data using body structure and physics rules to pin down real-world metric measurements, without needing 3D training data. Tested on running, track and field, and figure skating, it proved very accurate. It stayed within 33-41 mm of vertical error and ran smoothly right on an iPhone 15.
Imagine watching someone run and trying to pinpoint the exact spot where their weight is balanced, but you're looking at them through a single security camera. Normally, a flat video doesn't show depth, so you can't tell if an arm is reaching forward or just sideways, making it hard to find that balance point. Here, the system first finds where each body part is in the video, then separately estimates how far away they are, and finally uses basic facts about human anatomy to calculate the true center of mass in three dimensions.
Coaches and rehabilitation specialists can track an athlete's weight distribution during movement using just a smartphone, making biomechanical analysis accessible outside expensive labs and improving training feedback.
The system is only evaluated on specific athletic movements like running and figure skating, so it may not generalize to other activities. It requires a one-time calibration step, and the depth updates occur asynchronously at a lower rate than pose estimation.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
MuyBridge: Mobile Human Center-of-Mass Estimation from Monocular Video via Sparse Fusion — Aidan Bradshaw, Marco Giordano, David Rode et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.