Creating with style – style transfer
Another way we can assist an artistic team is via style transfer, a process that involves combining two images:
- The style image or root image, from which we will learn the style
- The target image, which we will transform with the new style
The resulting image will retain the core elements of the target image but appear to be painted or printed following the style image.
There are several methods for performing style transfer, including leveraging GANs (described in the previous section), using Visual Geometry Group (VGG), and employing Stable Diffusion (which we will cover in the next section).
In style_transfer.ipynb, we will use VGG19, a special type of CNN with 19 layers that has been trained with over a million images from the ImageNet database to extract the style of a Picasso painting and transfer it to a photograph. Picasso belonged to the cubism movement, where the artists applied multiple perspectives, used...
