Development of an Algorithm with Computer Vision using YOLOv8 Neural Networks for Blueberry Quality Inspection

Abstract

The article presents the development of an Algorithm with Computer Vision using YOLOv8 Neural Networks for Blueberry Quality Inspection. The YOLOv8 network was employed for the detection and classification of blueberries based on their quality. A total of 840 images of blueberries were collected and labeled using the Roboflow platform. After training and evaluating the model, an accuracy ranging from 89% to 96%, and F1-Scores between 90% and 97% were achieved in classifying blueberries as good or bad across seven different production zones. The results demonstrate the effectiveness of the YOLOv8-based computer vision system for accurately detecting blueberry quality, optimizing the selection process, and reducing human intervention.