Seeing as I’m starting to get more serious about machine learning, I’ve created this website to document my adventures. Most of my previous posts have been in rather impermanent media, so my first few posts will simply be playing catchup with what I’ve done in the past. This post specifically will be dedicated to the Neural Style algorithm, specifically this implementation, as I’ve found it to be very reliable.
Future posts will be much more sporadic, with one or a few images at a time. If you have a request for a content/style pair, I’ll process those when I can and post the result if it’s something I think could be cool.
What is neural style?
The Neural Style algorithm first appeared shortly after google’s Inceptionism algorithm (also commonly known as Deep Dream) first hit the public. The purpose of the algorithm is to separate the style of an image from the content, and then try to create an image which optimally matches the style of one image and the content of another.
What have I done?
Lots! This is one of my favorite applications of neural networks, short of some state-of-the art generative models I haven’t managed to get my hands on quite yet. Here’s a collection of my favorite results and various historically significant results.