<![CDATA[Waste management is an important challenge in protecting the environment and public health. Improperly managed waste can cause pollution and hinder the recycling process. This study aims to classify waste based on images by optimizing three deep learning architectures, namely EfficientNetB3, MobileNetV2, and ResNet50, to determine the model with the best performance. The dataset comes from the Kaggle platform, consisting of 4,650 images in six categories: battery, glass, metal, organic, paper, and plastic. The research stages include preprocessing, data augmentation, model development, and evaluation using accuracy, precision, recall, and F1-score metrics. The results show that EfficientNetB3 with the Adam optimizer achieved the best performance with 93% accuracy, followed by ResNet50 with 91%, while MobileNetV2 ranged from 70–73% depending on the optimizer. Variations in optimizers were found to affect model performance, while data augmentation improved generalization capabilities, especially in classes with limited samples. This research confirms the potential of deep learning methods in supporting automatic waste classification systems and provides a basis for the development of technology-based waste management systems in the future.]]>
Copyrights © 2025
