SMART ELECTRICITY CONSERVATION SYSTEM USING EFFICIENTNET

Abstract

Conservation of electric resource has been one of the important challenges over the decades. Worldwide, many nations are struggling to conserve and to bridge the gap between the demand and production of the resource. Though many measures like several Government acts, replacing existing products with energy conserving products and many solar based systems are being invented and used in practise, the demand and the need to preserve the resource still persists. Hence, this paper focuses on a novel technique to conserve the electric resource using a deep learning technique. The system uses Convolutional Neural Networks to identify and localize humans in the CCTV footages using EfficientNet, a deep transfer learning model. The classifier processes and yields its output to an embedded Arduino microcontroller, after detecting the presence/absence of human. The microcontroller enables/disables the electric power supply in the area of human’s presence/absence, based on the classifier’s output respectively. The system achieves an accuracy percentage of 84.2% in detecting humans in the footages with the subsequent enabling/disabling of electric power resource to conserve electricity.