Psalmantha Allaine Ipong, Maria Mae Kiskis, Angie Ceniza-Canillo
Milkfish (Chanos chanos) is one of the Philippines’ key aquaculture species, yet its post-harvest processes—especially sorting and weighing—remain mostly manual, making them slow and inconsistent. This study applies morphometric analysis and ensemble learning to automate size classification and weight estimation of milkfish using image data. Images collected by the researchers and a Roboflow dataset of 2,739 labeled samples were used. Ten morphometric features, including length, area, aspect ratio, and Feret diameter, were extracted. Random Forest and Gradient Boosting with Support Vector Machine as the meta-classifier were used for milkfish classification, while ensemble regression through weighted averaging was applied for weight prediction. The models achieved 95% accuracy, and a root mean square error of 139.60 grams, showing strong potential to streamline and standardize post-harvest operations in aquaculture. ©2025 IEEE.
Department of Computer, Information Sciences, and Mathematics, University of San Carlos, Cebu City, Philippines
