Garbage Classification Model for Enhanced Sorting and Categorization

Global waste production is an environmental problem that we face today. Every year there are 2.01 billion tons of waste produced. Apart from the increasing amount of waste, most of this waste is not managed properly. To enhance recycling and trash management procedures, efficient waste sorting is crucial. Deep learning models-more especially, Convolutional Neural Networks (CNN)-are used in this study. Using the TrashNet dataset which consists of 2,527 images divided into 6 types of trash. We train and evaluate 3 CNN models, namely DenseNet169, VGG19 and ResNet512. The model with the highest accuracy, DenseNet169, has an F1 Score of 0.9553 and an accuracy of 0.9575. The RenseNet512 model, which has an F1 Score of 0.9288 and an accuracy of 0.9327, comes next. Then the VGG19 model is in last place with an accuracy of 0.9169 and an F1 Score of 0.9148. This research shows the potential of CNN in increasing the accuracy of waste classification, which can create more sustainable waste management practices.