Title: Artificial Intelligence and Computer Vision Approaches for Automated Waste Classification: A Systematic Review
Authors: Fatuma S. Yusuf, Irine Leonard, Ada Ludigija, Daniel Mfungo
Volume: 10
Issue: 7
Pages: 659-668
Publication Date: 2026/07/28
Abstract:
Rapid urbanization and rising municipal solid waste generation have made source segregation a central challenge for cities in developing regions, where mixed-waste disposal undermines recycling, causes economic loss, and creates environmental and public-health hazards. Artificial Intelligence (AI) and Computer Vision, particularly Convolutional Neural Networks (CNNs) and transfer learning, have emerged as promising tools for automating waste identification from digital images. This paper presents a systematic review of empirical research on AI-based waste classification, tracing its evolution from early sensor- and feature-based machine learning methods, through the deep learning and transfer learning era, to recent hybrid, ensemble, and multi-modal approaches. The review synthesizes the data-related, model-related, and deployment-related challenges reported across studies, together with the augmentation, transfer learning, ensemble, compression, and domain-adaptation strategies proposed to address them. Five research gaps are identified: limited research in East African contexts, weak integration between model development and end-user applications, the absence of standardized real-world benchmarks, insufficient study of user trust, and a shortage of models optimized for low-resource hardware. The review concludes by outlining a design-science research direction, illustrated through an ongoing project in Iringa, Tanzania, that responds directly to these gaps.