Plataforma de Monitoramento Comportamental Baseado em Visão Computacional para Apoio ao Acompanhamento Pedagógico no Ensino Básico

Abstract

This work presents the development of a classroom behavioral monitoring platform based on Artificial Intelligence and Computer Vision techniques. The proposed solution integrates facial recognition using the InsightFace library and posture analysis through the YOLOv11-Pose model, enabling the automatic identification of observable student behaviors in real time. The platform is capable of classifying behavioral patterns relevant to the educational context, such as attention, distraction, participation, agitation, and drowsiness, based on the analysis of images captured in the classroom environment. The collected data are processed and organized in a structured database, allowing longitudinal monitoring and visualization through an interactive interface developed in Streamlit. The main objective of this study is to support pedagogical analysis by assisting teachers in identifying behavioral patterns and understanding classroom dynamics during lessons. The system is designed as a functional proof of concept and does not aim to perform clinical diagnoses, but rather to provide objective insights to support educational practice. The results demonstrate the technical feasibility of the proposed solution and its potential as a tool to support pedagogical decision-making, while respecting the ethical and interpretative limitations associated with the use of computer vision technologies in educational environments.