Virtual Booth

This paper introduces a system that combines attendance logging and locker opening using AI-driven facial recognition. The system uses a custom-trained MobileNet V2 convolutional neural network (CNN) model with TensorFlow Lite for quick and accurate facial recognition on a Raspberry Pi platform. We also developed facial pre-processing techniques to enhance image clarity, which improves recognition accuracy. The facial recognition system is integrated into a smart school setup. When a student is identified, their attendance is logged in an online database, and a push button unlocks their locker. This dual functionality saves significant time for taking attendance and reduces the burden of heavy bags for students by providing lockers. Our results show that the AI model achieves 99% accuracy during training, and around 92% accuracy in real-world testing on the embedded platform. In terms of speed, the system can recognize a face in about 300 milliseconds, log attendance in 800 milliseconds, and open a locker in 1300 milliseconds. With its relatively low development cost, this system is affordable and practical to be used at both primary and secondary schools.