A Smart Classroom Based on Deep Learning and Osmotic IoT Computing

Abstract

The biggest growth rate of network traffic in the coming years will be for smartphones and Internet-connected devices, which relentless tend to perform increasingly demanding tasks on continuously increasing amounts of data. Machine Learning and Edge Computing are emerging as effective paradigms for processing huge amounts of data supplied by the Internet of Things and Smart Cities. An osmotic computing architecture for an IoT smart classroom is used for testing a deep learning model for person recognition. A comparative performance study and analysis was made by means of selecting a single deep learning model, that it was tried to be adapted to run over the cloud, a fog microserver and a mobile edge computing device. The results obtained shown some promising results and also limitations for the edge and fog computing side that will need to be addressed in order to minimize latencies and achieve real-time responses for the present IoT application.