Thermoplastic waste segregation classification system using deep learning techniques

This research proposes a deep learning-based system, named deep CNN architecture, for the automated classification of the plastic resin in plastic waste. The system aims to detect and recognize objects such as drinking water bottles, detergent bottles, squeezable bottles, and plastic plates, and segregate them into PET, PE-HD, PE-LD, and other resin categories. The process involves capturing input images through a camera and using deep learning or traditional algorithms to detect and recognize the objects by comparing them with a trained database containing labeled objects. Unrecognized objects are dynamically trained, labeled, and updated in the database. The proposed system is implemented using Python, a versatile open-source programming language. Python's functional and aspect-oriented programming paradigms are leveraged to develop the models. The performance of the proposed architecture is evaluated against existing works, demonstrating a classification accuracy of 92.66% according to experimental results.