Implementation of Style Transfer Using the VGG-19 Network Architecture

VGG-19 is a CNN consisting of 19 layers. It was trained using millions of images from the ImageNet database. The network is capable of classifying images into 1,000 different class labels, including a vast number of animals and different tools.

Note

To explore the ImageNet database, go to the following URL: http://www.image-net.org/.

Considering its depth, the network is able to identify complex features from a wide variety of images, which makes it particularly good for style transfer problems, where feature extraction is crucial at different stages and for different purposes.

This section will focus on how to use the pretrained VGG-19 model to perform style transfer. The end goal of this chapter will be to take an image of an animal or a landscape (as the content image) and one of a painting from a well-known artist (as the style image) to create a new image of a regular object with an artistic...