Abstract
This article presents an innovative artificial intelligence-based system developed for the real-time monitoring of children’s emotional states within preschool settings. The system employs a deep convolutional neural network (CNN) specifically designed to recognize emotions through the analysis of facial expressions. The FER-2013 dataset served as the foundational data source for training a model that categorizes emotions into seven distinct groups. Throughout the training and validation phases, the model achieved an accuracy rate of 81.32%. Results generated by the system are disseminated to educators and parents via a web application built on the Django framework, as well as through a Telegram bot interface. The proposed methodology enhances the continuous observation of children’s psycho-emotional states in kindergartens, promotes timely intervention strategies, and fosters improved communication between parents and educational institutions.