A Deep Learning Aided Smart Waste Classification System for Smart Cities

Waste management is one the major global challenge today to ensure a sustainable environment, are some of the events that have assumed importance courtesy of the growing temperatures as well as global warming. The main aim of this paper is to identify waste images of different categories; these are labelled as paper, plastic, glass, metal, and organic materials. Accordingly, this paper proves that the identification of waste and their classification is efficient and can be executed via deep learning strategies with Convolution Neural Network (CNN) architecture. The comparison is made with an ensemble learning method known as Extreme Gradient Boosting (XGBoost). A VGG16 CNN network with total layers that include batch normalization, flattening, dense and dropout are used for waste classification and assessing them. The metrics that show the learning curve values obtained from the training of the model with the datasets from SENTINEL-2 exhibit a fairly good accuracy for both the validation set and the training set. These results validate that the XGBoost model and its applicability in the waste classification mechanism and are efficient when compared to other machine learning models.