Grabalut: Automated Robotic Arm System for Duck Egg Detection and Classification Using YOLOv8

Balut is a popular Filipino delicacy typically sold in the streets. Traditionally, balut is manually sorted from penoy (unfertilized or underdeveloped duck eggs) using the candling method, which is considered labor-intensive and prone to human error and misclassification. To address this challenge, this study presents an AI-powered robotic arm system capable of classifying and sorting duck eggs into either balut or penoy. The system integrated YOLOv8 for computer vision-based classification and a 6-degree-of-freedom (6DOF) robotic arm with a soft gripper for automated egg handling and sorting. Through detailed methodology involving data collection, annotation, preprocessing, model training, and validation, the study demonstrated a high accuracy with a mean average precision (mAP) of 0.96. Results showed that the system achieved a precision of 95.72%, a recall of 100%, and a F1-score of 97.81% using YOLOv8 trained with epoch 100. These metrics highlighted the effectiveness of the system in accurately detecting and classifying the duck eggs. The findings of the study showcase the potential of AI-driven computer vision and robotic automation in modernizing the traditional egg sorting process, enhancing efficiency and productivity in the balut industry.