Attendance System

Our Team

• Ammar Yasser    (Team member)  AIU
• Jamal Khaled     (Team member)  AIU
• Seif Elkerdany     (Team leader)     AIU
• Shrouq Waleed    (Team member)  AIU

Features

• Universal Course Integration: The system is built to support attendance tracking for any course offered at the university, providing a scalable and adaptable solution.
• Seamless Student Enrollment: Enrolling students is straightforward—just upload one or more reference images per student. The system will automatically use these for future attendance verification.
• Camera-Based Face Recognition: Attendance is recorded by simply scanning students' faces in real-time as they look at the camera, eliminating the need for manual input or cards.
• Intelligent Face Matching: The system uses advanced face verification techniques to compare live-captured images with the enrolled database and mark attendance accurately.

Project Outline

• 1. Data collection & annotation

  • 1.1 Collecting Student faces.

    • Different face reactions.
    • Different light conditions.
    • Every student must submit from 2 to 5 images.

  • 1.2 Explore the Data

    • Know more about the data.
    • The number of images.
    • The images formats.
    • The brightness of the images.
    • The sharpness and the quality of the images.

  • 1.3 Face detection and cropping

    • Using MTCNN to detect and crop faces.

  • 1.4 Data preprocessing

    • Resize images.
    • Remove noise from images.
    • Normalize the brightness.

  • 1.5 Generate pairs for training

    • Organize images in a directory.
    • Split the dataset into train, test and validation.
    • Create PyTorch dataset class.

• 2. Neural Network

  • 2.1 Model architecture

    • Implement the Siamese Neural Network with modifications.

  • 2.2 Data augmentation

    • Horizontal flip
    • Random rotation
    • Change brightness
    • Zoom out some images
    • Changing contrast

  • 2.3 Training loop

    • Loss function
    • LR schedular
    • Early stopping and save the best weights

  • 2.4 Evaluating the model

    • Keep updating and changing the model until it reaches great accuracy.

• 3. Attendance checking

  • 3.1 Classroom image processing pipeline

  • 3.2 One-to-many face matching

    • Create enrollment directory

  • 3.3 Attendance logging

  • 3.4 Save attendance records

    • Save in CSV file format / xlsx file format

• 4. Deploying as a website with a database.

Future updates

• Edge Device Optimization: Enhance the system’s efficiency to run seamlessly on edge devices such as Raspberry Pi and NVIDIA Jetson, enabling offline and portable deployment.
• Real-Time Video Processing: Extend functionality to support real-time video streams for continuous face detection and attendance logging, rather than relying solely on still images.
• Accuracy Enhancement: Improve the face verification model’s accuracy and robustness to meet high standards.