LUMINA Vision Node — Alzheimer AI Monitor

LUMINA Vision Node is an AI-powered patient safety system for Alzheimer and dementia patients. 55 million people worldwide live with Alzheimer's — families ask daily: "Is my loved one safe?" LUMINA answers automatically, in real time. CORE TECHNOLOGY: LUMINA uses a hybrid computer vision pipeline. Primary detection: face recognition via dlib-based 128-dimensional encodings. Fallback: clothing color fingerprinting using HSV histogram back-projection — tracks the patient even when face is not visible. This dual-layer approach minimizes false negatives without requiring wearable devices. KEY FEATURES: - Live CCTV Monitoring — MJPEG stream at 25 FPS with real-time bounding box overlay - Face Recognition — Patient confirmed via face encoding; prevents false positives from guests - Color Fingerprint Fallback — HSV back-projection tracks patient when face is not visible - GPS Safe Zone — Haversine-formula geofencing alerts family when patient leaves home radius - Emotion Detection — Automatic inference: calm, anxious, or confused from activity patterns - SOS Emergency — Single-tap emergency with GPS coordinates sent to family dashboard - Memory Gallery — Family uploads photos to help patient recall loved ones - Stage Adaptation — AI reports adapt to Alzheimer stage (1=Mild, 2=Moderate, 3=Severe) AMD INTEGRATION: LUMINA uses a provider-based architecture for vision inference. The system routes snapshot analysis to AMD Developer Cloud running LLaVA vision model via vLLM and ROCm. Switching to AMD hardware requires only updating two environment variables (AMD_VISION_URL, AMD_MODEL_NAME) — zero code changes needed. Architecture is deployment-ready for AMD MI300X with 192GB HBM3. IMPACT: Built for 160 million family caregivers worldwide who cannot afford 24/7 professional care. LUMINA works with existing CCTV — no new hardware required. All patient data stored locally for maximum privacy. Because dignity should not depend on wealth.