Artificial Intelligence By Example

Next convolution and pooling layer

The next two layers of the CNN repeat the same method as the first two described previously, and it is implemented as follows in the source code:

# Adding a second convolutional layerprint("Step 3a Convolution")classifier.add(Conv2D(32, (3, 3), activation = 'relu'))imp=classifier.inputoutputs = [layer.output for layer in classifier.layers]print("Step 3b Pooling")classifier.add(MaxPooling2D(pool_size = (2, 2)))

These two layers have drastically downsized the input to 14 x 14 x 32, as shown in this diagram: 

Convolution and pooling layer

The next layer can now flatten the output.