Applied Deep Learning - Part 4: Convolutional Neural Networks

The result of this process is an input image where the filter is very active. Remember that each filter acts as a detector for a particular feature. The input image we generate will contain a lot of these features. We will visualize filters at the last layer of each convolution block. To clear any confusion, in the previous section we visualized the feature maps, the output of the convolution operation. Now we are visualizing the filters, the main structure used in the convolution operation. We will visualize 8 filters per layer.