Enhanced security and attendance with a modified FaceNet and facebox detector for high-speed edge deployments.
This project addresses the challenge of traditional attendance tracking by implementing a facial recognition system that offers enhanced security and efficiency.
Using a modified FaceNet architecture and an optimized facebox detector, our solution achieves real-time recognition with high accuracy even in varying lighting conditions.
The system was specifically designed for edge deployment, allowing institutions to process data locally without internet dependency, maintaining privacy and reducing latency.
Implemented custom pre-processing pipeline with adaptive histogram equalization and noise reduction techniques.
Optimized model through quantization and pruning, reducing the model size by 70% while maintaining 98% of accuracy.
Designed a system that stores only feature vectors rather than actual facial images, and implemented local processing to avoid cloud dependencies.
"The facial recognition attendance system has revolutionized how we track student attendance, saving us countless hours and improving accuracy significantly."
Dr. Sarah Johnson
Dean of Technology, State University
