📘 Smart Attendance System (AI + ML + MERN + DevOps)

An AI-powered smart attendance system that uses Real-time Face Recognition to automatically detect a person and store attendance in a secure database. This project demonstrates Machine Learning, Full-Stack Development, Microservices, DevOps, Docker, and Cloud deployment concepts.

• 🎯 Real-time Face Recognition using Python + OpenCV + face_recognition

• 🤖 ML Microservice exposed as a Flask REST API

• 🔐 Liveness Detection (Anti-Spoofing)

  • Blink detection (EAR algorithm)
  • Micro head movement detection
  • Prevents cheating via photos/screens

• 🗄️ Backend API built with Node.js + Express

• 💾 Attendance stored in MongoDB

• ⚡ React Frontend Dashboard (real-time updates)

• 🐳 Containerized with Docker

• ☸️ Deployable on Kubernetes

• 🔄 CI/CD using GitHub Actions

To prevent users from cheating the system using passport photos, printed photos, or mobile screens, the ML model now includes Liveness Detection.

Attendance is stored ONLY if a real human is detected.

• Detects natural blinking using facial landmarks
• Fake images/screens do not blink
• If EAR < 0.20, blink detected → Real human

• Detects left/right micro head movement
• Photos/screens remain perfectly still
• If nose keypoints shift > 5px, movement detected → Real human

If (blink detected OR head movement detected):
      → Allow face recognition + save attendance
Else:
      → Return "Spoof Detected" (no attendance saved)

+---------------------+        +----------------------+        +---------------------+
|  Face Recognition   | -----> |   Node.js Backend    | -----> |     MongoDB         |
|  (Python + Flask)   |        |   (API + Mongoose)   |        |  (Database Storage) |
+---------------------+        +----------------------+        +---------------------+
                                     ↑
                                     |
                              React Frontend
                           (Live Attendance View)

attendance/
│
├── backend/               → Node.js + Express + MongoDB REST API
│   ├── src/
│   │   ├── models/
│   │   ├── routes/
│   │   ├── app.js
│   │   └── server.js
│   ├── .env
│   └── package.json
│
├── ml/                    → Machine Learning microservice
│   ├── known_faces/
│   ├── encodings.pkl
│   ├── app.py             → Flask API + Liveness + Recognition
│   ├── recognizer.py      → Face matching + backend communication
│   ├── liveness.py        → Blink + head-movement detection
│   ├── encode_faces.py    → Converts images → encodings.pkl
│   └── requirements.txt
│
├── fronted/ (or frontend) → React.js dashboard
│   └── src/App.js
│
└── README.md

• Python
• OpenCV
• face_recognition library
• Flask API

• Node.js
• Express.js
• MongoDB
• Mongoose

• React.js
• Axios
• Modern UI Table

A docker-compose.yml configures all three services plus MongoDB. Make sure Docker Desktop (or Docker Engine) is running.

1. Build & start everything

   docker compose up --build

   • frontend → http://localhost:3000
   • backend → http://localhost:8000
   • ml service → http://localhost:5000
   • MongoDB → mongodb://localhost:27017/attendance

2. Stop / remove containers

   docker compose down

⚠️ The ML container requires camera access. On Linux use --device /dev/video0:/dev/video0 or run in privileged mode. Windows/macOS need extra setup (e.g., use host as the video source or send frames via API).

cd attendance/ml
pip install -r requirements.txt
python encode_faces.py
python app.py

ML runs on 👉 http://127.0.0.1:5000/detect

cd attendance/backend
npm install

Add .env:

MONGO_URI=mongodb://localhost:27017/attendance
PORT=8000

Start backend:

node src/server.js

cd attendance/fronted
npm install
npm start

Runs on: 👉 http://localhost:3000

Alok Ranjan Smart Attendance System — AI + ML + MERN + DevOps