Desarrollo de un Algoritmo con Visión Artificial utilizando Redes Neuronales Convolucionales para el Control de la Calidad del Arándano

The paper presents the development of a computer vision algorithm using convolutional neural networks for blueberry quality control. The YOLOv8 network was used for the detection and classification of blueberries according to their quality. A total of 840 blueberry images were collected and labeled using the Roboflow platform. After training and evaluating the model, an accuracy between 91% and 98%, and F1-Scores between 90% and 97% were obtained in classifying blueberries as good or bad in seven different production areas. The results demonstrate the effectiveness of the YOLOv8-based machine vision system for accurate detection of blueberry quality, optimizing the sorting process and reducing human intervention.