Creative AI meetup #18: Style Transfer

In April, we will look at the creative potential of style transfer, a technology that uses the style of one image to draw the content of another. Marco Marchesi (Happy Finish) will talk about his experience using style transfer for a VR project for the Royal Academy, while David Pfau (DeepMind) will explore the neuroscience of style transfer and the limitations of the narrow definition of "style".

The event is curated by Luba Elliott and is part of the Creative AI meetup series designed to bring together artists, developers, designers, technologists and industry professionals to discuss the applications of artificial intelligence in the creative industries. To be notified about future events, sign up to the meetup and newsletter.

---

Marco Marchesi, Head of Technology, Happy Finish Ltd

"Can you make this image more neoclassical?" Practical uses of Style Transfer in the creative industry

Since the seminal work by Gatys, Style Transfer has become one of the most popular applications of Convolutional Neural Networks in the creative field. Despite its long sequence of successful results as an artistic tool, is still in its infancy in commercial applications, because of its limitations in the quality and size of the images that can be generated. But things are progressing fast, Marco will show a recent work made for the Royal Academy where Style Transfer was part of the CGI pipeline used to create a Virtual Reality experience, together with other recent use cases where Style Transfer is reaching unprecedented results.

Marco Marchesi is Head of Technology at Happy Finish. After an award winning career in Media Production and Visual Effects he focused his attention on the engineering side of the creative industry. After his research work in Human-Computer Interfaces and Computer Vision at the University of Bologna in Italy and in Conversational Design and Machine Learning at Apple Siri in California, he pursued a PhD in advanced technologies for Mixed Reality. In 2016 he joined Happy Finish to lead the development team responsible for the creation of Virtual Reality, Augmented Reality and Artificial Intelligence applications. Currently Marco is focused on implementing Deep Learning techniques into the company creative pipelines and in mixing realtime AR and AI on mobile devices.

----

David Pfau, Research Scientist at DeepMind

"The Neuroscience of Style Transfer"

Neural style transfer became an instant viral hit when it appeared two years ago. While it was mostly adopted as an "Instagram filter on steroids," few people are aware that it was originally developed as a tool for neuroscientists to understand visual cognition. I will tell the story behind neural style transfer, explain what it can tell us about how we perceive the world, and also highlight the limitations of the narrow definition of "style" in neural style transfer. Extrapolating from these limitations, we can speculate about how the development of AI in the 21st century may mirror the development of art in the 20th century, as we are able to develop progressively less superficial and more conceptual definitions of "style" that reflect a deeper understanding of our own minds.

David Pfau is a research scientist at DeepMind. Prior to that, he earned a PhD in computational neuroscience from Columbia University, focusing on the development of machine learning algorithms for neural decoding, particularly with an emphasis on Bayesian methods. His research at DeepMind is broadly focused on unsupervised learning, whether for generative models of images, texture synthesis and style transfer, reinforcement learning or neuroscience.

----

The schedule for the evening will be as follows.

6.30pm - 7pm Arrive

7pm - 7.10pm Introduction

7.10pm-7.50pm First talk - Marco Marchesi

7.50pm-8.30pm Second talk - David Pfau

#LDNcreativeAI

@elluba