What is markerless motion capture, and how accurate is it?

Markerless motion capture estimates the positions of body landmarks directly from video using computer-vision pose estimation, with nothing attached to the body. A 2024 systematic review and meta-analysis reported good to excellent agreement with marker-based systems for spatiotemporal parameters and walking speed, with weaker agreement for joint angles outside the sagittal plane and at the ankle.

Last reviewed .

How it works

A conventional marker-based system tracks reflective markers physically attached to the body. A markerless system removes the markers and infers the same landmarks from the image.

  1. Video is recorded of the person performing the task.
  2. A pose estimation model locates anatomical key points — joint centres and body landmarks — in each frame.
  3. A spatial reference, such as a calibration object or known camera geometry, converts image coordinates into real-world distances.
  4. Key point trajectories over time yield gait events, spatiotemporal parameters, and estimated joint angles.

Systems differ in how many cameras they use. Multi-camera arrangements reconstruct movement in three dimensions; single-camera approaches work from one two-dimensional view, which is more practical but constrains what can be recovered.

What the published evidence supports

A 2024 systematic review and meta-analysis in Sensors compared markerless camera-based systems against marker-based systems in gait analysis, including 22 studies. It reported overall good to excellent accuracy, validity, and reliability for spatiotemporal parameters, described accuracy and concurrent validity for walking speed as excellent with only a small bias, and found good-to-excellent intraclass correlation coefficients of 0.81 to 0.98 in the meta-analysis of walking speed, step time, and step length.

For joint kinematics, the same review found moderate to excellent agreement at the hip and knee, and poor concurrent validity and reliability at the ankle.

A study in stroke survivors comparing a markerless system against an instrumented walkway reported good to excellent agreement for spatiotemporal parameters, while stride width and paretic single-limb support time showed poor agreement.

Where the method is weakest

  • Frontal and transverse plane joint angles, where agreement with marker-based systems is substantially poorer than in the sagittal plane.
  • Ankle kinematics, consistently the least reliable joint across studies.
  • Narrow lateral measures such as step and stride width, which depend on resolving small distances.
  • Discrete range-of-motion values, which have shown more variability than continuous joint angle waveforms.
  • Population coverage — much validation work has been conducted in healthy adults, and reviews note limited attention to patient populations.

Reviews also emphasize that methodological procedures are not standardized across markerless systems, which limits how far a finding about one system transfers to another.

The findings above describe the specific systems evaluated in those studies. They are not performance claims for CurveCapture™ or CurveResearch™, and no equivalence between CurveAssure and any system in that literature should be inferred.

Why it is used anyway

Marker-based motion capture is costly, time-consuming, requires a dedicated laboratory, and needs a skilled operator to place markers correctly. Those constraints, rather than any deficiency in accuracy, are what keep it out of routine care.

For spatiotemporal parameters — the measures most functional outcome work actually relies on — the published agreement between markerless and marker-based systems is strong enough that the practical constraints dominate the choice.

Where CurveAssure fits

CurveAssure is a markerless, video-based system. A calibration floor marker provides the spatial reference, computer vision estimates key points on the body, and processing happens in the encrypted cloud rather than on the recording device. Nothing is attached to the patient.

Device descriptions covering the computer vision motion analysis system, the calibration floor marker, and data handling are available to authorized research partners.

References