Exposes webcam H-FOV and camera height offset as tunable sliders
in settings. Adds a subtle floor grid to the AR scene for spatial
reference.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Compute proper vertical FOV from assumed 63° horizontal (was incorrectly
using 63° as vertical, making depth ~2x off)
- Dynamically set camera height from worldLandmarks leg length instead of
hardcoded 1.1m — adapts to actual person proportions
- Pass worldLandmarks from MediaPipe result to AR scene
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace left-leg-raise confirmation with stand-still detection.
User walks to center, stops moving, and a 3-2-1 countdown confirms
the position. Uses EMA-smoothed ankle positions to detect stillness.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace jump-to-lock with a positioning flow: after calibration,
user walks to center of play zone and raises left leg to confirm.
That foot position becomes the locked AR square center.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Raycast foot landmarks onto a virtual ground plane via Three.js
transparent WebGL canvas. A sci-fi radial grid follows the feet
until the person stands still for 3 seconds, then locks as the
world center.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
name