A Design of Novel Method for Classification of Waste Materials with its location using Deep Learning and Computer Vision for Smart Cities

Abstract

Waste management is the one of the main  problems in all over the globe, presently waste materials  are collected and sorted by hand, it is very time consuming  and it also requires so much man power. Improper  management of waste materials leads to hazards including  environmental deterioration, soil contamination, water  pollution, and air pollution. To solve this problem, there  may be a requirement for an automatic method to aid to  recognize the type of waste substances and it’s Position.  Today’s technology is so sophisticated due of Artificial  Intelligence and Machine Learning. These technologies  may be utilized to address various real time issues, this  article handles the fundamental challenge of detecting and  separating the waste items like plastic, paper and metal  with their location. In this article above stated issue is  addressed using the Faster RCNN (Region Based  Convolutional Neural Networks) model which is very  much accurate compared to other algorithms like YOLO  (You Look Only Once) and other similar algorithms. The  model is trained on a custom dataset gathered on a mobile  camera and pre-processed using Label-Img Tool. Data  collected with different light conditions and in unique  angles. The model is trained using Faster R-CNN identify  objects and to obtain locations. This may assist individuals  to keep their surroundings tidy and to become conscious of  the garbage substances and to identify them. This paper has  been precisely recognizing kind of items and places with  higher accuracy.