Waste management is still a significant environmental problem, especially in developing countries with high population density. The waste sorting process, which is still mostly done manually, causes low efficiency and consistency of sorting. Previous studies generally focused on binary classification, while research on lightweight CNN architectures for multi-class waste classification, particularly in simulation-based scenarios with per-class performance analysis, remained relatively limited. This study aims to evaluate the performance of the waste image classification model using a lightweight Convolutional Neural Network architecture, namely MobilNetV2 based on the transfer learning approach. The dataset used consists of four categories of waste (metal, plastic, paper, organic) obtained from public datasets. The flow of this research includes the dataset collection stage, preprocessing in the form of resize and rescale, training model and evaluation model using accuracy, confusion matrix, precision, recall, and F1-Score metrics. The test results showed that the model achieved an accuracy of 92.02% with relatively consistent performance in each class. The misclassification that appears is limited and shows a similar pattern between classes. These findings show that MobileNetV2 is able to provide stable classification performance and has the potential to be used as a supporting model in the development of automated waste sorting systems in the future.
Copyrights © 2026