Desenvolvimento de um Sistema de Reconhecimento Facial para Controle de Acesso e Segurança no Campus IFPE Paulista.

Abstract

This work addresses the growing relevance of facial recognition for security in educational institutions, motivated by recent episodes of school violence, and proposes an access control system for IFPE's Paulista campus, whose methodology includes a literature review, selection of deep learning algorithms (Res10-SSD for detection and FaceNet for embeddings), and implementation in Python with OpenCV and face_recognition for real-time operation. The prototype processed 120 faces at rates between 17.44 and 31 FPS, with average confidence above the defined threshold. This allowed it to correctly identify authorized users and intruders, although performance fluctuations under load and lighting variations reveal the need for optimizations in pre-processing and diversification of the dataset.