CACAO BEANS CLASSIFIER USING CONVOLUTIONAL NEURAL NETWORK (Kakaw-Suri)

Abstract

This study presents the development of "Kakaw-Suri," a classifier designed to evaluate cacao beans based on their internal characteristics using Convolutional Neural Networks (CNN). The Philippines national standards for cacao bean grading prioritize internal characteristics, typically assessed via the cut test. However, CSU Lasam Cacao Processing Center in Lasam, Cagayan, revealed that local cacao processing centers often rely on external characteristics, overlooking internal defects such as mold, slate, and insect damage. Addressing this gap, the study focuses on classifying cacao beans internally, aligning with national standards to enhance produce quality. Using a Raspberry Pi, C920 Logitech camera, and an LCD screen, the researchers utilized YOLOv5, a CNN model, for training on 469 images. The researchers then developed a custom kiosk to house the components and implemented a GUI. The classifier demonstrated a notable performance, achieving a 94.39% accuracy rate during testing at the Cagayan Valley Cacao Development Center, Isabela State University. Additionally, the system was evaluated by respondents from Lasam, Cagayan, using ISO 25010 and ISO 9241-210 standards, achieving overall mean scores of 4.07 and 4.08, respectively. These results indicate that Kakaw-Suri can reliably classify cacao beans by internal characteristics, ensuring compliance with Philippine national standards and potentially improving the quality and marketability of local cacao produce.