Waste patterns based on deep learning object detection YOLOv8 in tourism areas

The increasing population growth and demand for disposable goods and waste production complicate the sorting and processing of hazardous, inorganic, and organic waste. Meanwhile, various waste processing techniques are needed for various types of waste, including unsafe, inorganic, and organic. This research aims to address the challenges of an inefficient waste management system by utilizing deep learning technology to help better classify waste. The contribution of the research is to use lightweight deep learning to learn waste types and obtain models. The method used is YOLOv8, a lightweight object detection algorithm for classification so that it is hoped that it can help manage waste types. The advanced architecture of YOLOv8 and its integration with frameworks such as TensorFlow and PyTorch facilitate accurate and efficient waste detection. The YOLOv8 architecture is used because it can detect objects based on frames. The dataset includes styrofoam, cardboard boxes, plastic bottles, cans, and plastic wrappers. Based on the research results, the average model accuracy was 96%, with an average error value of MSE 0.0065, RMSE 0.0806, and MAE 0.0025. The training and model creation process took ten minutes. The model was tested using experimental data with an accuracy confidence level of 85-95%. This research shows that YOLOv8 can improve waste management in the area