CNN-Based Cloud System for Student Identification from Incomplete Registration Records

Abstract

Universities need a decent student identification system to avoid unauthorized attendance. Nonetheless, missing information in registration records creates serious challenges for current recognition schemes. We created a cloud-based facial recognition system using ResNet-based convolutional neural networks to recognize students' faces from video-based images with incomplete enrollment data. The system utilized 270 images taken from nine students under different conditions; using these images, a result of 97% accuracy was achieved at a similarity of 75% to 100%. The approach utilizes notifications when unauthorized access is detected and uses cloud infrastructure for deployment. These results show that identification in these situations can be done using deep-learning models. This can prove fruitful in campus security and other registration management services.