Facial Recognition Attendance System using DeepFace - Blackcoffer

Summarize

This project was done by the Blackcoffer Team, a Global IT Consulting firm.

Contact Details
This solution was designed and developed by Blackcoffer Team
Here are my contact details:
Firm Name: Blackcoffer Pvt. Ltd.
Firm Address: 4/2, E-Extension, Shaym Vihar Phase 1, New Delhi 110043
Skype: asbidyarthy
WhatsApp: [phone]
Telegram: @asbidyarthy

Project Title
Facial Recognition Attendance System

The Problem
Manual attendance tracking is time-consuming and prone to errors such as buddy punching (proxy attendance). Existing biometric systems (like fingerprint scanners) require physical contact, which may be inconvenient, especially in environments where hygiene is essential.

Our Solution
The Facial Recognition Attendance System provides a touchless, efficient, and accurate way of marking attendance. The system uses a webcam to capture the user's face, compare it against a pre-registered database, and automatically mark attendance if a match is found. This eliminates manual tracking and ensures authenticity.

Solution Architecture
User Registration:
Collect user details (name, email) and capture facial images.
Store user information and face embeddings in MongoDB.
Mark Attendance:
Capture the user’s face using the camera.
Compare the captured face against registered embeddings using cosine similarity.
If a match is found, attendance is marked in the database with a timestamp.
Database Management:
MongoDB stores user details and logs attendance records.
UI:
Developed with Streamlit for an easy-to-use interface.
Users can navigate between Register and Mark Attendance sections through a sidebar.

Deliverables
Web Application:
A user-friendly interface for registration and attendance.
Database Integration:
Stores user details and attendance logs in MongoDB.
Facial Recognition Engine:
Uses DeepFace to verify the user's identity from captured images.

Tech Stack
Tools Used
Streamlit: For building the web interface
MongoDB: For storing user data and attendance logs
DeepFace: For facial recognition
NumPy: For performing similarity computations
Pillow (PIL): For image processing
Language/Techniques Used
Python: Core language for development
Facial Recognition Models: Facenet model through DeepFace
Models Used
Facenet Model: Extracts 128-dimensional face embeddings for recognition
Skills Used
Web development
Facial recognition
Database management
Data processing and similarity matching
Databases Used
MongoDB: NoSQL database to store users and attendance
Web Cloud Servers Used
Local Server (via Streamlit)
MongoDB Atlas for cloud database management

What are the Technical Challenges Faced during Project Execution
Handling Face Variations:
Faces with/without glasses or changes in lighting affected recognition.
Performance with Multiple Users:
As the number of registered users grew, the recognition process slowed.
Session Handling:
Ensuring that the registration and attendance pages behave independently without conflicts.
Accuracy Threshold Tuning:
Finding the right similarity threshold to balance false positives and negatives.

How the Technical Challenges were Solved
Face Variation Handling:
Enabled users to register multiple face images to improve recognition accuracy.
Performance Optimization:
Used cosine similarity for faster matching and optimized the way embeddings are retrieved from MongoDB.
Session Management Fixes:
Used Streamlit’s session state to manage page navigation and inputs.
Threshold Adjustment:
Fine-tuned the similarity threshold to accommodate slight variations while avoiding incorrect matches.

Business Impact
Increased Efficiency: Attendance marking is instantaneous and automated, saving time for both students/employees and administrators.
Reduced Errors: Eliminates human errors associated with manual attendance entry.
Contactless Solution: Ideal for environments with hygiene concerns, especially post-pandemic.
Enhanced Security: Prevents fraudulent attendance practices like proxy attendance.

Project Snapshots (Minimum 10 Pictures)
Home Page: Navigation menu for Register and Attendance sections.

Register Section: Form to input name, email, and capture image.

Image Capture: Camera input for user registration.

Registration Confirmation: Success message after saving user details.

Mark Attendance Page: Camera input for capturing face.

Attendance Matching: Success message if face matches a registered user.

Error Handling: Face not recognized message when no match is found.

MongoDB Database View: Sample of stored user data and attendance logs.

Log File: Screenshot of attendance records.
Performance Optimization: Loading logs showing system behavior with multiple users.

Project Website URL
This project is currently hosted locally. To run it, clone the repository and execute: