ID-Guard

Revolutionizing attendance with AI-driven face recognition, real-time analytics, and anti-spoofing security for seamless and reliable classroom management.

Comment

Inspiration

Attendance systems in educational institutions are often manual, time-consuming, and vulnerable to proxy attendance. This leads to inaccurate records and reduced accountability.

We wanted to build a system that is fast, reliable, and secure using AI. ID-Guard was created to eliminate manual errors and introduce a scalable, real-time attendance solution with built-in protection against spoofing.

What it does

ID-Guard is an AI-powered attendance management system that uses face recognition to automatically mark attendance in real time.

  • Detects and identifies students instantly
  • Marks attendance without manual intervention
  • Prevents spoofing using liveness detection
  • Provides real-time analytics and attendance trends
  • Sends automated email notifications for absentees

Unlike traditional systems, ID-Guard combines real-time face recognition with liveness detection, ensuring both automation and security.

The system verifies identity using embedding similarity:

$$ d(E_1, E_2) < \theta $$

How we built it

We built ID-Guard using a modern full-stack architecture:

  • Frontend: Next.js (React + TypeScript) with Tailwind CSS
  • Backend: FastAPI for real-time face recognition
  • AI/ML: Face detection and encoding using deep learning (128-D embeddings)
  • Liveness Detection: Flask microservice for anti-spoofing
  • Database & Storage: Firebase Firestore and Cloud Storage
  • Authentication: Firebase Auth with JWT
  • Notifications: Resend API for email alerts and Twilio for whatsapp notification

Face recognition pipeline:

$$ Image \rightarrow Detection \rightarrow Encoding \rightarrow Matching \rightarrow Decision $$

Challenges we ran into

  • Maintaining high accuracy under different lighting conditions
  • Preventing spoof attacks using photos or screens
  • Achieving real-time performance with low latency
  • Handling multiple face detections in a single frame

We optimized the system such that:

$$ T_{total} = T_{detect} + T_{encode} + T_{match} < 100 ms $$

Accomplishments that we're proud of

  • Achieved approximately 99.2% face recognition accuracy
  • Built a real-time system with response time under 100ms
  • Successfully implemented anti-spoofing using liveness detection
  • Designed a scalable cloud-based architecture
  • Developed a clean and modern analytics dashboard

What we learned

  • Practical implementation of computer vision and deep learning
  • Building and optimizing real-time AI systems
  • Integrating cloud services for scalability
  • Importance of security in biometric systems

We also learned how tuning system thresholds affects performance:

$$ Accuracy = \frac{1}{FP + FN} $$

What's next for ID-Guard

  • Mobile application (React Native)
  • Advanced analytics and ML-based predictions
  • Integration with institutional ERP systems
  • Multi-language support
  • Enhanced reporting and export features

Built With

Share this project:

Updates