OUR APPROACH
Useful observations.
Honest boundaries.
Strideglass separates what the software computes from what an AI says about it. Each reported observation includes a method, a usable video section, quality information, and the pipeline version.
What we inspect
The worker detects body landmarks and searches for uninterrupted walking sections. Where the footage supports it, the report shows candidate foot-contact timing, estimated cadence, and movement projected into the camera’s image.
Candidate contacts are estimated from landmark trajectories. They are not measured ground-contact forces. A movement seen in a two-dimensional image is not a calibrated three-dimensional joint angle.
What quality means here
Landmark visibility describes the model’s tracking signal. It is not the probability that a measurement is correct. Camera position, clothing, occlusion, motion blur, turns, and unusual movement can all affect the result.
We suppress measurements that fail individual checks. A report can show an inspectable contact sequence while withholding cadence or other estimates.
What we don’t conclude
This release is for adult walking education. It does not diagnose conditions, grade clinical severity, infer muscle weakness, analyze running, or prescribe corrective exercises based on unvalidated asymmetry.
The original Zeni study supports an event-detection method under its study conditions. It does not establish this phone-video implementation’s accuracy. Local synthetic and fixture testing are engineering checks, not population-level clinical validation.
Sources you can inspect
Two simple methods for determining gait events during treadmill and overground walking using kinematic data ↗
- Foot-marker movement relative to the pelvis can inform candidate gait-event timing.
Limits: The study used laboratory marker tracking and force-based comparison, not this monocular phone-video implementation. Do not transfer the study's reported accuracy to this service. A visually estimated event does not establish ground contact force.
Adult Activity: An Overview ↗
- Adults generally benefit from regular physical activity; some activity is better than none.
- The general adult guideline includes at least 150 minutes of moderate activity per week and muscle-strengthening activity on two days.
Limits: General education is not an individual walking prescription. These guidelines do not validate gait measurements or prescribe a target cadence.
Service recording and measurement protocol, version 1 ↗
- Use a fixed camera with the full body and both feet visible, steady lighting and a clear walking path.
- Record a side view for projected sagittal measurements; keep straight walking passes separate from turns.
- Cadence is estimated from accepted alternating events within a continuous usable segment.
Limits: This protocol is an engineering quality-control rule, not an independently validated clinical method. Image-plane movement changes with perspective and cannot establish true three-dimensional joint angles or muscle strength.
AI explanations
Your chosen provider receives structured observations, selected conversation context, and our curated source library. Raw video is not sent to AI providers. AI explanations can still be wrong; inspect the replay and cited sources when evaluating a claim.