Figure 2.

Best by accuracy/speed ratio neural network architectures. (a)—neural network with 2 convolutional layers and 3 intermediate fully connected layers; (b)—network with single convolutional layer and the stack of 7 small fully connected layers; (c)—network with 2 convolutional layers and 2 big fully connected layers (this model showed the best performance); (d)—model with 2 convolutional layers and the stack of 6 intermediate fully connected layers; CV—convolutional layers; MP—max-pooling layers; FC—fully connected layers.