Pose Estimation
Also known as: human pose detection, body keypoint detection
Pose estimation is the computer-vision process that detects the positions of major body joints (keypoints) in each video frame. This produces the skeleton that SwingVantage uses to measure angles and movement patterns.
Pose estimation runs a neural network over each video frame to locate landmarks like shoulders, elbows, wrists, hips, knees, and ankles. It is the first processing step after the quality gate and the foundation for all motion metrics. Accuracy depends on visibility: occluded joints (hidden by clothing, body rotation, or poor lighting) produce lower-confidence keypoints that are labeled accordingly. SwingVantage uses on-device pose estimation where possible to keep video private.
You don't need to understand how pose estimation works. This means just know that good video setup (light, angle, frame rate) directly improves its accuracy. This improves the reliability of everything else.
Review the skeleton overlay before trusting any metric. If a joint is visibly drifting off your body at a key moment (like impact), the metric derived from that joint at that frame should be treated as an estimate, regardless of the stated confidence.
Example
Pose estimation places 33 body landmarks on each frame of a golf swing, letting the engine compute hip-to-shoulder separation angle through impact.
Why it matters
Without reliable pose estimation, angles and timing metrics cannot be computed at all. It is the step where good video setup pays off most.
How it shows up on video
The skeleton overlay is the visible output of pose estimation. This means when estimation is working well, joints track smoothly. When it fails, landmarks jump or disappear, usually around fast-moving limbs or occluded joints.
Common mistakes
- Treating all pose estimation output as equally reliable — landmark confidence varies by joint, frame, and video quality.
- Assuming the system always knows where your wrist is during a fast swing — fast-moving distal joints are the hardest to track reliably.
- Wearing clothing that matches the background color, which causes landmarks to drift.
- Not checking the skeleton overlay at all — it is the most direct way to verify that tracking was accurate before trusting any analysis finding.
In SwingVantage Motion Lab
Motion Lab runs pose estimation on every submitted frame before building the 3D avatar. Frames where estimation confidence drops below threshold are interpolated or marked as gaps in the timeline scrubber.
Related terms
- Skeleton OverlayA skeleton overlay is the on-screen visualization of detected body joints and the lines connecting them, drawn over your video. As a result, you can see exactly what the system tracked.
- Landmark TrackingLandmark tracking follows the position of each detected body keypoint across consecutive video frames. This creates a time-series trajectory for every joint that enables timing and velocity measurements.
- Mobile Motion CaptureMobile motion capture is the process of using a smartphone camera and on-device pose estimation to track body movement in real time or from a recorded clip. This means no specialized hardware required.
- Video Quality ScoreA Video Quality Score is a pre-analysis rating (0–100) that tells you how usable a submitted clip is before pose estimation begins. This means catching bad angles, motion blur, or poor lighting early.
Put this into your swing
SwingVantage can spot this in your own swing — free to start.