Development of a Computer Vision Application for Mango (Mangifera Indica L.) Fruit Defect Detection Using YOLOv8 Architecture

This study investigates the utilization of YOLOv8 architecture to detect and categorize defects in ‘Carabao’ mangoes, using machine learning and object detection methods, the model was trained on a dataset comprising 2,160 annotated images, augmented to enhance performance. Results demonstrated high accuracy rates: 100% for black spot, 80% for brown spot, and 83.33% for mango scab in real time application. The model exhibited a mean Average Precision (mAP) of approximately 95% across IoU thresholds ranging from 0.50 to 0.95, with 70% at an IoU threshold of 0.50, highlighting its precision in object detection. Additionally, the average recall rate was an impressive 93^%. A custom-built controlled environment ensured consistent and detailed imaging, while software implementation integrated the YOLOv8 model into a Tkinter-based application for real-time image processing and defect classification. This study underscores the effectiveness of the YOLOv8 architecture in mango defect detection, particularly for appearance-based defect classification.