Waste Classification Using Transfer Learning and DenseNet

Abstract

Efficient waste management is an essential component of sustainable development, because using traditional methods can be time-consuming and hazardous to the environment. Automating trash categorization procedures through the incorporation of deep learning algorithms is a possible solution to these problems. This article suggests a garbage detection system that successfully classifies different waste kinds by using deep learning algorithms. Through the application of machine learning and deep learning, the system can discern between various waste categories, making appropriate sorting and disposal easier. The waste categorization efficacy of DenseNet architecture and transfer learning. The study uses pre-trained algorithms to improve sorting accuracy and refine them using a dedicated waste dataset. We investigate the possibility of DenseNet's dense connection patterns to extract complex characteristics and relevant patterns from waste photos. Significant accuracy gains are shown by the experimental findings, demonstrating the effectiveness of this strategy in waste classification system optimization. The experimental findings show significant accuracy gains, demonstrating this strategy's effectiveness in optimizing waste classification systems procedures. Through the creation of more precise and automated waste categorization systems, this research helps to promote environmental sustainability through the advancement of robust waste sorting technology. ResNet152, Inceptionv3, DenseNet were used in this work. The best accuracy comes from ResNet152, which is 98%