9 DeepDream and Neural Style Transfer

“Great things are done by a series of small things brought together.”

In fine art, especially painting, humans have mastered the skill to create 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 in creating art? In this chapter, we introduce an artificial system based on a Deep Neural Networks 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.

In this chapter, we are going to explore two new techniques to create artistic images using neural networks: 1) DeepDream and 2) Neural Style Transfer through the following three sections:

9.1   How convolutional neural networks see the world

9.1.1  Revisit how neural networks work

9.1.2   Visualize CNN features

9.1.3   Implement feature visualizer

9.1.4   Wrapping up

9.2   DeepDream

9.2.1   How does the DeepDream algorithm work?

9.2.2   DeepDream implementation in Keras

9.2.3   Wrapping up

9.3   Neural Style Transfer

9.3.1   The content loss

9.3.2   The style loss

9.3.3   Total variance loss

9.3.4   Network Training

9.3.5   Wrapping up