HYBRID DEEP LEARNING APPROACH USING YOLO V11 AND CNN FOR REAL-TIME APPLE OBJECT DETECTION AND RIPENESS CLASSIFICATION

Abstract

Apple farming, particularly in Indonesia, still faces numerous challenges because the adoption of modern technology for detecting fruit ripeness directly on the tree remains low. This information is crucial, as the correct harvest time heavily depends on it. To address this issue, this study aims to develop a system capable of detecting and classifying apple ripeness directly on the tree using a hybrid deep learning approach. We combine two advanced algorithms: YOLOv11 (You Only Look Once Version 8), utilized for rapid apple detection, and a Convolutional Neural Network (CNN), employed for classifying the apples as either ripe or unripe. This hybrid model is designed to maximize performance, as each model plays a distinct and complementary role. The system developed is implemented as a website-based application. The model was trained using a comprehensive dataset: 1,000 images of apple trees (for training YOLOv11) and 3,000 images of apples (ripe and unripe) for training the CNN model. The system yielded outstanding results, achieving a ripeness classification accuracy of 99%. This success demonstrates that this hybrid system has significant potential to be a practical solution for enhancing the efficiency and accuracy of harvest time determination, thereby supporting the modernization of the apple farming sector.

Kata kunci: Deteksi Object, YOLO V11, Convolutional Neural Network (CNN), deep learning, klasifikasi apel