Single Use Plastic Bottle Recognition and Classification Using Yolo V5 and V8 Architectures

Improper disposal of single use plastic bottles leads to many problems including danger to marine life and land pollution. Burning of plastic in turn releases dioxins and polychloride biphenyls. They are very harmful if inhaled and are threat to vegetation too. Manual sorting of plastic bottles and safe disposal is not an easy task. A lot of recycling initiatives use manual sorting for plastic recycling, which depends on plant staff visually identifying and selecting plastic bottles as they move along the conveyor belt. Automatic sorting of plastic bottles has advantage of non-intrusive sorting, speed, consistency, cost effectiveness in long run and even prevents health hazards to workers working in recycling environment. As a result, it is imperative to replace human sorting systems with intelligent automated systems. In this study, convolutional neural network architectures such as YOLOv5 and YOLOv8 were utilized to detect plastic bottles in images. Despite YOLOv8 having more parameters and requiring more computation time, it was found that YOLOv8 outperformed YOLOv5 in accurately identifying plastic bottles in the images.