Desarrollo de un Sistema de Visión Artificial con la Red Neuronal Convolucional (YOLO v8) para Clasificar el Arándano por su Grado de Madurez

In the context of precision agriculture, the use of computer vision systems has gained importance in optimizing the classification of agricultural products, especially blueberries. This study focuses on the implementation of a system based on the YOLOv8 convolutional neural network to classify blueberries according to their maturity level. The accuracy in classifying blueberry maturity is crucial to ensure their quality and optimize their commercial value. The methodology includes the collection and labeling of 666 images of blueberries in three stages of maturity: unripe, ripening, and ripe, using the Roboflow platform. Data augmentation techniques were applied to improve the variability of the dataset. The YOLOv8 model, trained in Google Colab with a Tesla T4 GPU, demonstrated high accuracy and efficiency in classification, achieving an accuracy of over 95%. The YOLOv8 architecture allows for simultaneous object detection and classification through convolutions and anchor layers, optimizing the evaluation of blueberry maturity. Despite the promising results of YOLOv8, further research is needed to expand its application. It is crucial to increase the diversity of the dataset and improve training to maintain accuracy under adverse conditions. Integrating YOLOv8 with other AI systems, adapting it to other crops, developing user-friendly interfaces, and conducting cost-benefit studies are essential. These improvements can enhance the system's efficiency and accuracy, promoting more sustainable and profitable agricultural practices.