Hands-On Reinforcement Learning with Python

CNN architecture

Now, let's just see how all of these layers are organized in the CNN architecture, as follows:

First, the image is passed to the convolutional layer, where we apply the convolution operation to extract features, and then feature maps are passed to the pooling layer where the dimensions are reduced. We can add any number of convolution and pooling layers depending on the use case. After this, we can add a neural network with one hidden layer at the end, which is known as a fully connected layer, which classifies the image.