Figure 7.

Typical architecture of a CNN. It takes the input image, suitably resized, and elaborates it through a series of internal layers consisting of convolutional ones, max-pooling layers, and fully connected layers until final classification^10–20.