Art & AI: The Logic Behind Deep Learning 'Style Transfer'

When humans and machines collaborate, we can produce things neither would create on their own. The intersection of art and AI is an area that I find really exciting, but with all the business impact AI can have, I personally feel it doesn't always get enough attention. In this spirit, I recently set out on a personal quest to learn more about PyTorch, the machine learning library that's been creating a lot of buzz since its 1.0 release late last year, and I was pleasantly surprised by what I found. For me, PyTorch turned out to be more than just an interesting alternative to TensorFlow, with dynamic graphs and an imperative coding style. One of the examples from their official docs inspired me to track down some academic papers and take a closer look at the inner-workings of a deep learning technique that's fascinated me for quite some time -- 'Neural Style Transfer'.