Classification of Organic and Non-Organic Waste Using Convolutional Neural Network (CNN)

Abstrak

The increase in waste volume in Indonesia, which reached emergency levels in 2024, requires technological solutions that can assist in the sorting process quickly and accurately. Previous research on CNN-based waste classification generally focused on recyclable waste categories with many classes and used structured datasets, which did not adequately represent real-world waste conditions, especially organic waste, which has more varied shapes and conditions. Based on this gap, this study proposes a Convolutional Neural Network (CNN) model for classifying two main categories—organic and inorganic—using 25,077 images and direct testing on field samples. The model was trained using the Adam optimizer and categorical crossentropy loss. The results show high accuracy for inorganic waste (96%), but lower accuracy for organic waste (62%) due to the complexity of texture and natural damage. This study contributes to the field of informatics through the application of more applicable and realistic deep learning for automatic waste sorting systems, as well as opening up opportunities for the development of model architectures that are more adaptive to waste conditions in the actual environment.