Style Transfer in PyTorch

Style transfer in machine learning is a technique that allows the artistic style of one image to be applied to another image. It involves extracting the style features from one image (e.g. colors, textures, patterns) and applying them to the content of another image while preserving its original structure. This process is often achieved using deep neural networks to separate and recombine the content and style representations of images.

In this project I use a pretrained convolutional neural network to create a style transfer application in PyTorch. The notebook for this project can be found as a static webpage here. The code is bassed on Gatys et al. and the PyTorch for Deep Learning and Computer Vision Udemy course from Slim et al.

Content image (left) is a phtoo of one of the many waterfalls in the Clemson, SC area. The style image (middle) is Edvard Munch's "The Scream", and on the right we have the final composite image.

Below is a gif showing some of the iterative steps of the style transfer.

Gif showing some stages of the image transfer application.