AI-based plastic waste classification for sorting purposes: A review on recent progresses and challenges

The rapid growth of plastic waste has heightened environmental concerns and created a pressing need for efficient classification for sorting purposes and recycling accordingly. In recent years, Artificial Intelligence (AI) based identification and classification, particularly Machine Learning (ML), Deep Learning (DL), and computer vision, has emerged as a revolution in automating the the process of identification and classification for sorting plastic waste. This systematic review, following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines, identified 112 articles published between 2015 and 2025 on AI-based plastic waste classification and sorting. This study critically reviewsthe current state-of-the-art classification and sorting methods, such as convolutional neural networks (CNNs), you only look once (YOLO) architectures, and Transformer-based models, and differentiates with AI-based classification for sorting purposes and assesses their integration with advanced spectroscopic techniques such as near-infrared (NIR), Fourier Transform Infrared Spectroscopy (FTIR), and Raman spectroscopy. The review highlights the accuracy of identification and classification methods for plastic waste and identifies key challenges, including limited real-world datasets, scalability issues, and environmental variability. As a novel contribution, this review consolidates performance metrics, maps the diversity of AI-based classification models, and suggests future research directions focusing on lightweight models, multi-sensor fusion, and edge-AI deployment. This research provides a valuable technical resource and strategic guide for researchers, engineers, and policymakers working towards sustainable and scalable AI-driven plastic waste classification.