<![CDATA[Waste management in Indonesia remains a challenge, particularly in monitoring waste volume and classifying waste types, which are generally still carried out manually and are therefore less efficient. This study aims to develop a Machine Learning-based Android application to estimate waste volume and detect waste types using Convolutional Neural Network (CNN) and You Only Look Once (YOLO) models. The research method includes image data collection and labeling, pre-processing, model training, evaluation, and model implementation into an Android application using TensorFlow Lite. The dataset used consists of 1,584 images for waste volume estimation and 3,168 images for waste type detection, which were divided into training, validation, and testing data. The test results show that the CNN model achieved an accuracy of 90 percent in estimating waste volume. Meanwhile, the YOLO model achieved a precision of 96.59 percent, recall of 97.51 percent, and mAP@50 of 98.39 percent in detecting waste types. Implementation on the Advan Tab 7 Android device showed that the models were successfully run in TFLite format, with a CNN model size of 12,919 KB and a YOLO model size of 11,994 KB, as well as application memory usage of 80.34 MB during operation. In addition, testing using external data outside the training and testing datasets showed that limitations still exist under certain conditions. Overall, the developed system has the potential to serve as a technology-based solution to support more efficient waste management.]]>
Copyrights © 2026
