Identifikasi Kematangan Buah Apel Menggunakan Algoritma YOLO

Abstract

The classification of fruit ripeness plays a vital role in the agricultural product processing industry. Manual sorting based on visual perception is often subjective and inconsistent. This research proposes an automatic detection and classification system for apple ripeness levels, namely unripe, half ripe, and ripe, using the YOLOv8n object detection algorithm. A dataset of 1,800 apple images was collected and annotated using YOLO format, then trained on a lightweight YOLOv8n model for 30 epochs. The evaluation results showed high performance, with mAP@0.5 of 0.975 and mAP@0.5:0.95 of 0.959. Class-wise, the model achieved F1-scores of 0.94 for unripe, 0.93 for half ripe, and 0.88 for ripe apples. The confusion matrix indicated that most misclassifications occurred between the ripe and half ripe classes, suggesting feature similarity. The model demonstrated accurate and efficient detection, making it suitable for real-time fruit sorting applications. Future work may explore data augmentation, deeper YOLO variants, or integration with IoT devices for deployment in agricultural environments.