9 DeepDream and neural style transfer

In fine art, especially painting, humans have mastered the skill of creating unique visual experiences through composing a complex interplay between the content and style of an image. So far, the algorithmic basis of this process is unknown, and there exists no artificial system with similar capabilities. Nowadays, deep neural networks have demonstrated great promise in many areas of visual perception such as object classification and detection. Why not try using deep neural networks to create art? In this chapter, we introduce an artificial system based on a deep neural network that creates artistic images of high perceptual quality. The system uses neural representations to separate and recombine content and style of arbitrary images, providing a neural algorithm for the creation of artistic images.

9.1 How convolutional neural networks see the world

9.1.1 Revisiting how neural networks work

9.1.2 Visualizing CNN features

9.1.3 Implementing a feature visualizer

9.2 DeepDream

9.2.1 How the DeepDream algorithm works

9.2.2 DeepDream implementation in Keras

9.3 Neural style transfer

9.3.1 Content loss

9.3.2 Style loss

9.3.3 Total variance loss

9.3.4 Network training

Summary