Universitas Dian Nuswantoro (UDINUS), through the participation of Lawi, Ansarullah; Dermawan, Aulia Agung; Sholikun; Reza, Ivan Muhammad; Saputra, Filmada Ocky; Handayani, Sri; Aranski, Alvendo Wahyu; Mursalim, Muhammad Khairul Naim; and Hasibuan, Zainal Arifin, conducted an activity focused on enhancing machine learning models for waste management. The activity addressed efficient waste management in relation to sustainability and urban planning amid increasing waste complexity.
The activity built on previous research that used real-time data collection through Internet of Things (IoT) devices. Its focus was improving the quality and diversity of the underlying dataset through data preprocessing and labeling. UDINUS developed a raw dataset comprising 10,046 images collected from multiple sources. Because manual bounding-box annotation was labor-intensive, 1,093 images, representing 10.9% of the raw dataset, were fully annotated and used to train and evaluate the YOLOv8 model. The remaining 8,953 raw images were preserved for future semi-supervised and active learning experiments using different machine learning models, including Random Forest and Support Vector Machine (SVM).
The activity provides a dataset and insights into data preprocessing for waste classification. The source states that diverse and well-preprocessed data significantly enhance model performance and lead to better classification accuracy. The activity therefore contributes to advancing waste management systems by providing a robust dataset and insights into data preprocessing. UDINUS presents the activity as a foundation for further research and practical applications in smart cities.