Practical Convolutional Neural Networks

Step 6 – Creating a CNN model

Now we have the input; that is, x_image is ready to feed to the convolutional layer. We formally create a convolutional layer, followed by the max pooling:

layer_conv1, weights_conv1 =  
    LayersConstructor.new_conv_layer(input=x_image, 
                   num_input_channels=num_channels, 
                   filter_size=filter_size1, 
                   num_filters=num_filters1, 
                   use_pooling=True) 

We must have the second convolutional layer, where the input is the first convolutional layer, layer_conv1, followed by the max pooling:

layer_conv2, weights_conv2 =  
    LayersConstructor.new_conv_layer(input=layer_conv1, 
                   num_input_channels=num_filters1, 
                   filter_size=filter_size2, 
                   num_filters=num_filters2, 
                   use_pooling=True) 

We now have the third convolutional layer where the input is ...