Classification of Plastic and Non-Plastic Wastes Using Mobile net and SVM

Abstract

In recent years, governments across the globe are keen to envisage the use of AI, Deep learning and machine learning methods with the objective of limiting the ill effects of improper waste segregation methods. The current state of technology makes it possible to create a system that can identify plastics and non-plastics automatically from its image. It is necessary to do a feature extraction method in order to take the distinct features of the plastic/non-plastic object that can define the properties of the object in order to provide a more accurate classification.

This study suggests an architecture that uses Mobilenet for feature extraction and the extracted features were classified using Support Vector machine classifier to separate waste items into plastic and non-plastic. Compared to a conventional CNN the proposed approach using Mobilenet requires less training parameters. The dataset employed in this work is a customized one exclusively compiled for this study. The proposed system of separating plastics from other materials requires less manual labour and can be used in smart garbage systems and for plastic segregation in industries.