Real anatomy. Quietly shown.

Replaced generic illustrative dots with the actual HNS landmark map your model tracks. Three opacity tiers signal density without overwhelming the user:

• 7 primary landmarks (bright, some live-pulsing): nose, chin, jugular notch, C7, L/R acromion, sternum top — the load-bearing measurements.
• ~30 secondary landmarks (35% opacity): face mesh (eyes, eyebrows, ears, jaw, cheeks, mouth), trap/clavicle line, sternum mid, ribs.
• 10+ micro mesh markers (18% opacity): forehead, cheek, shoulder, chest density signal.
• Connector lines at 25% opacity: centerline (head-to-jugular-to-sternum), shoulder line, collar diagonals, dashed ear-reference for tilt detection.
• "104 landmarks · live" tracking chip sits beneath the session pill — quiet credibility signal that the system is doing serious work without showing off.
• Behavior in production: primary landmarks pulse on detection confidence ≥ 95%. If a landmark is occluded or low-confidence, it fades. The user instinctively learns "PostureVision sees me" without ever being told.