Yield Quantification of Morus Serrata Fruit Stages Using YOLOv8-Based Identification

Morus Serrata is a nutritious and medicinally valuable fruit with notable ecological benefits. In the Philippines, it is widely cultivated by both farmers and hobbyists due to its zero-waste harvest process. Traditional methods for assessing mulberry ripeness are labor-intensive and timeconsuming. This study evaluates the effectiveness of an automated Jetson Nano-based system for multi-class detection and yield quantification of Morus serrata fruits, categorized into four stages: flowering, unripe, ripe, and overripe. The system incorporates a SONY IMX477 camera module for image acquisition. Performance is assessed using mean absolute error and correlation coefficient. Results show a strong alignment between predicted and actual fruit counts, with an overall correlation of 86.69 percent. The highest accuracy is observed in the flowering stage, while the unripe stage presents the most significant detection challenge. Despite its effectiveness, the system's performance is affected by the computational limitations of the Jetson Nano, especially during high-resolution processing. Future enhancements should focus on optimizing the processing pipeline, improving feature extraction, and utilizing more powerful hardware to achieve greater accuracy and realtime reliability.