Facial Recognition Security System

Your Identity is Your Access

Comment

ACM Facial Detection System - DevPost Submission

Inspiration

Many schools, companies, and organizations still use Excel sheets and manual records for attendance and access control. This is time-consuming and error-prone. We wanted to automate this process using facial recognition AI to help frontline workers (security guards, receptionists) do their jobs more efficiently—assisting them rather than replacing them.

Languages/Frameworks Used

  • Python, NumPy, Tkinter, dlib, OpenCV, face_recognition, SQLite3
  • Node.js, Express.js, JavaScript, HTML/CSS
  • CORS, Body-parser for API communication

What It Does

A complete facial recognition system for automated attendance tracking:

  • Real-time face detection from webcam
  • Automatic check-in/check-out logging with timestamps
  • Face registration interface for new users
  • Desktop app (Tkinter) and web dashboard (Node.js)
  • SQLite database storing user faces and attendance records
  • REST API for integration with other systems

How You Built It

  1. Facial Recognition Engine - Used pre-trained dlib models for face detection and encoding faces into 128-dimensional vectors
  2. Desktop Application - Built Tkinter GUI with OpenCV webcam stream for real-time detection and face registration
  3. Database Design - SQLite with two tables: one for storing face encodings, one for attendance records
  4. Web Backend - Express.js REST API for querying and managing attendance data
  5. System Integration - Created main.py orchestrator to launch both the Python app and Node.js server
  6. Image Processing - Converts frames, detects faces, generates encodings, and compares against known faces using distance metrics

Challenges You Ran Into

  • Different lighting conditions affecting recognition accuracy
  • Real-time processing speed on standard hardware
  • False positives (recognizing different people as the same)
  • Cross-platform GUI compatibility issues
  • Serializing NumPy arrays to SQLite database
  • Managing multiple processes (Node.js + Python) together
  • Privacy and consent concerns with facial data storage

Accomplishments You're Proud Of

  • Built a complete end-to-end system (not just a prototype) with desktop and web interfaces
  • Achieved real-time face recognition on standard hardware without GPU
  • Designed with ethics in mind—augmenting workers instead of replacing them
  • Works across macOS, Linux, and Windows
  • Clean, modular architecture that's easy to extend
  • Comprehensive documentation for deployment

What You Learned

  • How to deploy pre-trained deep learning models in real-world applications
  • Full-stack development (AI backend + web frontend)
  • Database design for complex data types (face embeddings)
  • Building systems with multiple coordinated processes
  • Computer vision fundamentals (image preprocessing, feature extraction, recognition)
  • Importance of ethical considerations in AI deployment
  • Cross-platform application development

What's Next for the Project

Near-term:

  • Multi-face detection (recognize multiple people at once)
  • Configurable detection sensitivity
  • Bulk user import from CSV
  • Enhanced reporting (daily/weekly/monthly summaries)

Medium-term:

  • GPU acceleration for faster processing
  • Liveness detection (prevent photo spoofing)
  • Email/SMS notifications for anomalies
  • Integration with HRIS systems (Workday, SuccessFactors)
  • Multi-camera support for larger spaces

Long-term:

  • Edge deployment with optimized models
  • Support for additional biometrics (iris, fingerprint)
  • Enterprise multi-site deployment
  • Advanced analytics and threat detection

Current Status: MVP complete and functional

Deployment Ready: Yes, with local-only architecture and privacy documentation

Built With

Share this project:

Updates