The increasing waste problem necessitates efficient solutions, one of which is automatic classification based on artificial intelligence. This study develops a Convolutional Neural Network (CNN) model for classifying organic and inorganic waste images using a transfer learning approach with the MobileNetV2 architecture. The model was trained in two stages, namely feature extraction and fine-tuning, using a dataset of 25,077 images from a public Kaggle repository. The results show that the model after fine-tuning achieved an accuracy of 92.28%, with a precision of 89.6% for the organic category and 96.4% for inorganic. High recall and F1-score values were also achieved, demonstrating that transfer learning with fine-tuning effectively improves waste image classification accuracy and has potential for implementation in automatic waste sorting systems.
Ardiansyah Putra (Wed,) studied this question.