Prototype Model For Plastic Bottle Waste Detection Using Yolov8s

Abstrak

Plastic bottle waste is one of the most common types of inorganic waste found in the environment and has become a serious environmental problem. Along with the rapid development of artificial intelligence technology, object detection methods based on deep learning can be utilised to automatically detect plastic bottle waste in digital images. This study aims to develop a prototype model for plastic bottle waste detection using the YOLOv8s algorithm. The research method used is descriptive quantitative with several stages including dataset collection, preprocessing, model training, testing, and evaluation. The dataset used consisted of 1001 images of plastic bottle waste obtained independently using a smartphone camera and additional datasets from Kaggle. The dataset was labelled and processed using Roboflow with image resizing to 512×512 pixels and divided into training, validation, and testing datasets. Model training was carried out using the Ultralytics library on Google Colab with parameters of 5 epochs, image size 512×512, and batch size 8. The results showed that the YOLOv8s model was able to detect plastic bottle objects properly in various environmental conditions, backgrounds, and lighting variations. Evaluation results indicated that the model achieved good performance based on precision, recall, mAP50, and mAP50-95 values. Therefore, YOLOv8s can be effectively implemented as a lightweight and real-time object detection model for plastic bottle waste detection.