<![CDATA[Waste management is a critical issue in urban areas due to increasing volumes and diverse waste conditions. In Bandung City, plastic and can waste with intact or dented states often complicate manual sorting, which is time-consuming and error-prone. This study proposes an automatic classification solution using Convolutional Neural Network (CNN) with a MobileNetV2 transfer learning approach. The dataset was obtained from Kaggle and preprocessed through normalization and resizing before training. Experimental results achieved 84.33% accuracy, with the best performance in metal classes (precision and recall above 87%) and the lowest in dented plastic (recall 66.67%). The model was integrated into a Streamlit-based interface for real-time prediction. These findings highlight CNN’s effectiveness in supporting faster and more consistent waste classification, although further dataset expansion is needed to improve performance in specific categories.]]>
Copyrights © 2025
