Deteksi dan Klasifikasi Kue Tradisional Indonesia Menggunakan YOLOv8

Abstract

Indonesian traditional cakes are part of the cultural heritage, characterized by their rich flavors, unique forms, and significant historical value. However, the lack of recognition among younger generations necessitates a new approach to preservation efforts. This study aims to develop an image processing-based detection system for traditional cake types using the YOLOv8 algorithm. The five types of cakes identified in this research are lumpur cake, lapis cake, wingko cake, dadar gulung cake, and putu ayu cake. The image dataset was obtained through a combination of direct image capture and public datasets, and was manually annotated using the Roboflow platform. The model was trained using the PyTorch framework and evaluated based on precision, recall, F1-score, and mean Average Precision (mAP) metrics. Experimental results show that the system achieved an average mAP of 89.9% and an F1-score of 86.5%, with a relatively low classification error rate. These findings indicate that the YOLOv8 algorithm is effective in detecting visually similar objects and holds significant potential for application in the digital preservation of culinary heritage. The system can also be further developed as a technology-based educational medium to support the conservation of Indonesia’s local culinary wealth.

Keywords: YOLOv8, Object Detection, Cake Traditional, Image Processing, Computer Vision