Development of a Computer Vision Application for Copra Meat Cooking Detection and Classification Using YOLOv8 Architecture

Traditional quality assessment methods of copra cooking are labor-intensive and subjective, underscoring the need for automated solutions. This study introduces a copra meat detection and classification application using the YOLOv8 architecture. The hardware setup comprises a custom-built box equipped with a high-definition camera. The software employs YOLOv8 integrated into a Tkinter-based application for real-time image processing. The model achieved an accuracy of 84.40% in live testing, as evaluated using a confusion matrix, indicating robust performance despite slightly lower accuracy in detecting over-cooked copra. Accuracy rates for each class were also determined: under-cooked copra achieved 90%, perfectly-cooked copra reached 86.67%, and over-cooked copra achieved 76.67%.