Algorithm Based on Deep Learning Techniques for Classification of Solid Waste in Recycling Plants

The imperatives of environmental sustainability and resource efficiency necessitate advancing recycling technologies, among which solid waste classification is a pivotal process. Traditional manual sorting methods, hindered by inefficiency, cost, and scalability issues, have given way to innovative solutions employing deep learning algorithms and specialized software, promising to revolutionize the solid waste management sector. This manuscript explores the burgeoning domain of algorithm-based classification systems for solid waste, specifically focusing on their application within recycling plants. It comprehensively studies these systems' effectiveness, precision, and environmental impact while examining their implementation's particularities in various contexts, including the Peruvian districts. Comparative analysis of CNN architectures suggests that while ResNet18 was sufficient, Inception-ResNet could yield higher accuracy due to its complexity and depth if computational resources were not a limiting factor.