ExamGuard AI is a multi-tier, AI-assisted examination monitoring and proctoring platform. It analyzes real-time video feeds and session metrics to generate an explainable risk score and structured evidence timeline for examiner review without making unilateral decisions.

• Multi-Role Support: Student exam portal, Examiner live monitoring dashboard, and Admin system management.
• AI/CV Microservice:
  • FACE_MISSING: Absence detection with persistence threshold.
  • MULTIPLE_FACES: Detection of additional persons in frame.
  • PHONE_DETECTED: Object detection for unauthorized mobile devices & materials.
  • EXCESSIVE_LOOKING_AWAY: Facial landmark & head orientation tracking.
• Real-Time Proctoring & Risk Engine:
  • Configurable event weights and cooldown periods.
  • Live alerts streamed to examiners via WebSockets (Socket.IO).
  • Snapshot & metadata evidence archiving.
• Exam Engine:
  • Question bank management and randomized question delivery.
  • Server-side countdown timers and answer auto-saving.

ExamGuard AI
├── frontend/          # React + Vite + Tailwind CSS (Student Exam UI & Examiner Dashboard)
├── backend/           # Node.js + Express + Socket.IO + Mongoose (Auth, Exams, Sessions, Risk)
├── cv-service/        # Python + FastAPI + OpenCV + YOLO (Real-time Vision Analytics)
└── extension/         # Manifest V3 browser extension for focus monitoring (Optional)

• Node.js (v18+)
• Python (3.10+)
• MongoDB (running locally or MongoDB Atlas URI)

(Detailed setup instructions will be updated as modules are developed)

MIT License.