<![CDATA[This research addresses the escalating global issue of waste management, particularly in developing cities like Tasikmalaya, Indonesia. Despite a daily waste production reaching 320 tons, the current management system remains manual and reactive. This study aims to transform this system into a proactive model by implementing Digital Image Analysis and a Convolutional Neural Network (CNN) algorithm to predict waste volume from images. The methodology involved the collection and augmentation of a waste image dataset, yielding a total of 3,168 images. Furthermore, a sequential CNN architecture was designed and trained over 50 epochs. The primary novelty and finding of this research lie in the developed CNN model, which achieved a high overall accuracy of 90% in volume classification. This performance demonstrates that computer vision provides an effective solution, significantly outperforming the 77.6% accuracy reported in a previous related study. Ultimately, this achievement marks a crucial step toward realizing a Smart Waste Management System. It establishes a data-driven foundation for optimizing collection schedules and resource allocation, despite minor challenges in distinguishing between highly similar volume classes (e.g., 90% and 100%).]]>
Copyrights © 2026
