Implementation of Yolov8 For Digital Image-Based Classification of Mangosteen Fruit Ripeness Stages

Implementasi Yolov8 Untuk Klasifikasi Tingkat Kematangan Buah Manggis Berbasis Citra Digital

Abstrak

Manual determination of mangosteen (Garcinia mangostana) ripeness still relies on human visual observation, which may lead to subjective and inconsistent assessments. This study aimed to develop a YOLOv8-based model for detecting mangosteen ripeness stages. The research adopted the Cross Industry Standard Process for Data Mining (CRISP-DM) framework, which includes business understanding, data understanding, data preparation, modeling, evaluation, and deployment stages. The dataset consisted of 600 mangosteen images classified into six ripeness stages, namely Stage 1 to Stage 6, with 100 images in each class. Image annotation was performed using bounding boxes, and the dataset was divided into training, validation, and testing subsets using a 70:20:10 ratio. Five YOLOv8 variants, namely YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, and YOLOv8xl, were trained and compared based on precision, recall, mAP50, and mAP50-95. The results showed that YOLOv8m achieved the best performance, with a precision of 0.772, recall of 0.941, mAP50 of 0.933, and mAP50-95 of 0.919. YOLOv8m was selected as the best model because it obtained the highest mAP and recall values, indicating its ability to detect mangosteen ripeness stages more effectively. These findings demonstrate that YOLOv8 has the potential to be used as a supporting system for automatic mangosteen sorting based on ripeness stage.