Featured Project Posts
AI-Based Face Recognition Attendance System using Deep Learning
Automated attendance system using CNN and FaceNet deep learning model. 99.2% accuracy on real-time video feed with Python and OpenCV.
Read Full Project
Smart Home Automation System using ESP32 and Google Assistant
Complete IoT smart home with voice control, mobile app, energy monitoring and Firebase cloud integration using ESP32 microcontroller.
Read Full Project
6-DOF Robotic Arm with AI Vision Guidance using SolidWorks & Python
Industrial robotic arm design with 6 degrees of freedom, AI-based object detection guidance, servo control and SolidWorks simulation.
Read Full ProjectAI-Based Face Recognition Attendance System using Deep Learning
Project Demo Video
Full demo video showing real-time face detection and attendance marking.
Project Abstract
This project presents an AI-powered automatic attendance marking system using Deep Learning-based face recognition technology. The system captures real-time video from a camera, detects faces using MTCNN (Multi-task Cascaded Convolutional Networks), and recognizes individual students using FaceNet embeddings and a Support Vector Machine (SVM) classifier. Recognized students are automatically marked present in a CSV/Excel database, eliminating manual attendance and reducing proxy attendance by 99%.
Problem Statement
Traditional manual attendance systems are time-consuming, error-prone, and susceptible to proxy attendance fraud. Students waste valuable class time during roll call, and attendance data is difficult to analyze. There is a critical need for an automated, biometric-based attendance system that can accurately identify students in real-time and maintain digital records automatically.
Proposed Solution
We propose a deep learning-based face recognition system using MTCNN for face detection and FaceNet for generating 128-dimensional face embeddings. An SVM classifier trained on student face images identifies individuals with 99.2% accuracy. The system integrates with a web dashboard (Flask) for real-time monitoring, attendance reports, and admin management. Data is stored in SQLite and exportable to Excel.
Hardware Requirements
Software Requirements
Key Features
- Real-time face detection & recognition
- 99.2% face recognition accuracy
- Multi-face detection simultaneously
- Automatic Excel/CSV attendance log
- Web dashboard with admin panel
- Student registration module
- Attendance reports & analytics
- SMS/Email notification support
- Proxy attendance prevention
- Mobile responsive web interface
Advantages
Applications
Technologies Used
Keywords / Tags
Detailed Project Explanation
The system operates in three phases: Training, Enrollment, and Recognition.
Phase 1 – Enrollment: Students register by providing their name and student ID. The webcam captures 100 face images per student. MTCNN detects and crops face regions, which are passed through FaceNet to generate 128-D embeddings stored in the database.
Phase 2 – Training: An SVM classifier is trained on the stored face embeddings. Data augmentation (rotation, brightness variation) improves robustness. The trained model achieves 99.2% cross-validation accuracy on test data.
Phase 3 – Real-time Recognition: During class, the webcam stream is processed frame-by-frame. MTCNN detects faces, FaceNet generates embeddings, and the SVM predicts the identity. Recognized students are marked present with timestamp in SQLite. The Flask web dashboard shows live attendance and generates Excel reports.
Get This Project
Share This Project
Related Projects