Streamlined representation of the architecture designed for the CNN model used in this work. Orange and blue items describe two different block operations, respectively: (i) convolution and activation function, (ii) convolution, activation function and pooling. The simultaneous reduction of the square dimensions and their increasing amount intuitively represent the abstraction process typical of a CNN. Green circular units are arranged in order to describe the fully connected (i.e. dense) layers. The dimensions of the feature maps are reported for each pooling operation, together with the number of features extracted by the CNN.