About the Project

This project presents a real-time smart attendance system that uses facial recognition to automate classroom attendance. By leveraging deep learning and deploying the system on the NVIDIA Jetson Nano, the proposed solution enables fast, accurate, and efficient attendance tracking. The system eliminates manual effort and reduces errors associated with traditional attendance methods.

Problem Statement

Traditional attendance systems such as manual roll calls, QR codes, and RFID-based methods suffer from several limitations:

• Time-consuming processes

• Possibility of proxy attendance

• Dependence on additional devices

• Lack of real-time monitoring

There is a need for a reliable, automated, and intelligent attendance system that ensures accurate identification and works efficiently in real classroom environments.

Existing Systems and Limitations

QR-Based Systems

• Requires smartphone for scanning

• Proxy attendance possible

• Slow for large classrooms

RFID-Based Systems

• Requires physical cards

• Cards can be lost or shared

• One-by-one process

• These systems lack automation and real-time intelligence.

Key Features

• Real-time face detection and recognition

• Automated attendance logging

• High accuracy and reliability

• Eliminates proxy attendance

• Works efficiently for large classrooms

• Edge-based processing (no cloud dependency)

Methodology

The system follows a structured pipeline:

• Data Acquisition: Classroom images captured via camera

• Preprocessing: Face detection, alignment, normalization

• Model Training: CNN-based deep learning model

• Recognition: Matching faces with stored database

• Output: Attendance generation and reporting

System Workflow

(Add your figures here from paper)

• Classroom image capture

• Face extraction using MTCNN

• Feature extraction and classification

• Attendance list generation

Results

The system demonstrates strong real-time performance:

• Efficient detection and recognition of student faces

• Accurate attendance generation

• Works under varying classroom conditions

• Robust to lighting and pose variations

Sample outputs include:

• Detected student faces

• Recognized identities

• Generated attendance list

System Performance

• Inference Time: ~0.8 seconds per frame

• Real-Time Capability: Yes

• Scalability: Suitable for 20–40 students

• Efficiency: Low power, edge computing

System Implementation

Hardware:

• NVIDIA Jetson Nano

• Camera Module

• Power Supply

• Wi-Fi Module

Software:

• Python

• OpenCV

• TensorFlow / PyTorch