A Systematic Review of CNN Architectures and Evaluation Metrics in Solid Waste Classification Using Deep Learning
A Systematic Review of CNN Architectures and Evaluation Metrics in Solid Waste Classification Using Deep Learning

Summary/Abstract: In recent years, both academia and industry have identified solid waste management (SWM) as a relatively new area of study. This paradigm shift has precipitated the emergence of advanced Machine Learning (ML) and Deep Learning (DL) applications that prioritize cost reduction, expedited processes, and automation, thereby minimizing human intervention. The objective of this study is to conduct a systematic review of the literature to identify the architectures and metrics used for the efficient classification of solid waste (SW). The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were utilized for the dissemination of the results. Consequently, a total of 41 articles were identified as relevant sources. The investigation revealed that the most prevalent Convolutional Neural Network (CNN) is VGG-16, which attains a classification accuracy of 96.1% within 100 epochs. A conspicuous absence in the extant literature is the application of R-CNN. The efficacy of this algorithm in the classification of solid waste is notable, as it facilitates the simultaneous categorization of multiple waste types.

• Details
• Contents