Распознавание изображений элементов зерновых смесей методами глубокого обучения с использованием библиотек keras и tensorflow

Abstract

The possibility of using deep learning neural networks for recognizing images of elements of grain mixtures was investigated using the method of computational experiment. Comparison of results for three different architectures - VGG16, VGG19 and MobileNet on a set of images of grains of various plants was made. The relative contribution of morphological, color and textural features to the solution of the problem of classification of grain images with the help of deep learning neural networks is revealed.