The LeNet architecture is one of the classic architectures of a CNN. As shown in the following diagram, the architecture is very simple, and it consists of only seven layers. Out of these seven layers, there are three convolutional layers, two pooling layers, one fully connected layer, and one output layer. It uses a 5 x 5 convolution with a stride of 1, and uses average pooling. What is 5 x 5 convolution? It implies we are performing a convolution operation with a 5 x 5 filter.

As shown in the following diagram, LeNet consists of three convolutional layers (C1, C3, C5), two pooling layers (S2, S4), one fully connected layer (F6), and one output layer (OUTPUT), and each convolutional layer is followed by a pooling layer: