🎓 Auto Attendance System

AI-Powered Facial Recognition Attendance for Modern Classrooms

---

A production-ready web application that automates classroom attendance using deep-learning face recognition. Capture a photo, detect every face in the frame, and instantly mark attendance — all in real time.

---

🧠 How It Works

┌─────────────┐     ┌──────────────┐     ┌───────────────────┐     ┌────────────────┐
│  📸 Camera  │────▶│  YuNet ONNX  │────▶│  ArcFace (ONNX)   │────▶│ 🔍 Cosine      │
│  Capture    │     │  Detection   │     │  512-D Embedding   │     │    Similarity   │
└─────────────┘     └──────────────┘     └───────────────────┘     └────────┬───────┘
                                                                            │
                         ┌──────────────────────────────────────────────────┘
                         ▼
               ┌───────────────────┐     ┌─────────────────┐     ┌─────────────────┐
               │ 🎯 Match Against  │────▶│ ✅ Mark          │────▶│ 📡 SignalR       │
               │    Student DB     │     │    Attendance     │     │    Broadcast     │
               └───────────────────┘     └─────────────────┘     └─────────────────┘

1. Capture — A photo is taken from a browser webcam or an RTSP IP camera.
2. Detect — The YuNet ONNX model locates every face in the frame.
3. Embed — Each cropped face is passed through ArcFace (InsightFace) to produce a 512-dimensional embedding.
4. Match — Embeddings are compared against stored student embeddings using Cosine Similarity (threshold ≥ 0.65).
5. Record — Attendance is saved to the database with duplicate-prevention logic.
6. Broadcast — SignalR pushes the result instantly to every connected dashboard.

---

✨ Features

Category	Feature
🤖 AI Engine	ArcFace (InsightFace) 512-D embeddings via ONNX Runtime
👁️ Face Detection	YuNet ONNX — fast, multi-face, rotation-robust
⚡ Real-time Updates	SignalR WebSocket broadcasts for live attendance feed
🔐 Security	ASP.NET Core Identity · Role-based access · CSRF protection
📊 Reports	Filter by Classroom / Faculty / Date Range · CSV export
📹 Capture Sources	Browser webcam + RTSP IP camera support
🧑‍🎓 Student CRUD	Full student profile management with multi-photo upload
🔄 Auto Training	Background model retraining when new photos are added
🛡️ Duplicate Guard	Configurable time window to prevent re-marking

---

🛠️ Technology Stack

Layer	Technology
Framework	ASP.NET Core 8.0 MVC (C#)
Database	PostgreSQL · Entity Framework Core 8
Face Detection	YuNet ONNX model
Face Recognition	ArcFace (InsightFace) ONNX model via Microsoft.ML.OnnxRuntime
Image Processing	OpenCvSharp4 (.NET wrapper for OpenCV)
Real-time	ASP.NET Core SignalR
Auth	ASP.NET Core Identity
Frontend	Bootstrap 5 · Custom glassmorphism CSS · JavaScript
Version Control	Git · Git LFS (for ONNX model files)

---

� Project Structure

Auto-Attendance-System-ASP.NET/
├── readme.md                       ← You are here
├── Report.pdf                      ← Project report
└── DemoAttendanceSystem/
    ├── .gitignore
    ├── .gitattributes               ← Git LFS tracking rules
    ├── README.md                    ← Technical documentation
    └── DemoAAS/
        ├── Controllers/
        │   ├── AttendanceController.cs    ← Capture, recognize, mark
        │   ├── StudentsController.cs      ← Student CRUD + photo upload
        │   └── HomeController.cs          ← Landing page
        ├── Services/
        │   ├── FacialRecognitionService.cs ← Core recognition pipeline
        │   └── ArcFaceEmbeddingService.cs  ← ONNX inference wrapper
        ├── Hubs/
        │   └── AttendanceHub.cs           ← SignalR real-time hub
        ├── Models/
        │   ├── Student.cs
        │   ├── StudentPhoto.cs            ← Includes FaceEmbedding field
        │   └── Attendance.cs
        ├── Data/
        │   └── ApplicationDbContext.cs
        ├── Views/                         ← Razor views (MVC)
        ├── arcface.onnx                   ← ArcFace model (Git LFS)
        ├── face_detection_yunet.onnx      ← YuNet model (Git LFS)
        └── Program.cs

---

🚀 Getting Started

Prerequisites

Requirement	Version
.NET SDK	8.0+
PostgreSQL	14+
Docker	24.0+ (optional)
Docker Compose	2.0+ (optional)

| Git LFS | 3.0+ (for cloning ONNX models) |

Installation

# 1. Install Git LFS (required for ONNX model files)
git lfs install

# 2. Clone the repository
git clone https://github.com/Daku3011/Auto-Attendance-System-ASP.NET.git
cd Auto-Attendance-System-ASP.NET/DemoAttendanceSystem

3. Configure the database — Edit DemoAAS/appsettings.json:

"ConnectionStrings": {
  "DefaultConnection": "Host=localhost;Database=DemoAAS;Username=postgres;Password=your_password"
}

# 4. Apply database migrations
dotnet ef database update --project DemoAAS

# 5. Run the application
dotnet run --project DemoAAS

🐳 Docker Deployment (Recommended)

Run the entire stack (App + Database) with one command:

# Build and start services
docker-compose up --build -d

# View logs
docker-compose logs -f

The application will be available at http://localhost:8080.

🌐 Open your browser at https://localhost:5001 (or the port shown in the terminal).

---

📋 Usage Guide

Step 1 — Register Students

Navigate to Students → Create New. Enter the student's details and upload 3–5 clear, front-facing photos per student. The system will automatically extract and store face embeddings.

Step 2 — Take Attendance

Go to the Attendance page. Click Start Camera, position students in the frame, and hit Capture & Mark Attendance. The system detects all faces, matches them, and logs attendance instantly.

Step 3 — Monitor in Real Time

Recognized students appear in the live sidebar via SignalR — no page refresh needed. Connected dashboards update automatically.

Step 4 — Export Reports

Visit Attendance Records → filter by Classroom, Faculty, or Date Range → click Export CSV.

---

🔮 Roadmap

• ArcFace (InsightFace) deep-learning embeddings
• SignalR real-time attendance broadcasts
• ASP.NET Core Identity authentication
• Continuous "Live Mode" scanning without manual capture
• Attendance analytics dashboard with charts
• Docker containerization for one-command deployment
• Mobile-responsive PWA for tablet kiosks

---