Desarrollo de un algoritmo de visión artificial con redes neuronales convolucionales (YOLO V8) para el control de calidad de la mandarina Murcott

In mandarin processing companies, the manual classification of fruits is carried out manually, this depends on human vision and the process does not follow a standardized selection criterion, so an artificial vision algorithm was designed with convolutional neural networks that improve the classification of mandarins by their state of ripeness by means of an algorithm in Python v3.7 and YOLO v8 when capturing the images with the webcam connected to the computer. It was possible to capture 2100 images, which were entered into a dataset of tangerines separating 1470 images for training, 420 for validation and 210 for neural network testing, obtaining 97.7% accuracy according to the confusion matrix in the classification of the fruit when tests were carried out in a controlled lighting scenario with real tangerines.