Skip to content
Self Portrait with Nolde,
Self Portraits by Leon Gatys

Leon Gatys

Self Portrait with Nolde, Self Portraits

Still, 2016

Buy Now $8000

Loading auction…

View full-resolution artwork ↗

About the artwork

From the very beginning, I used a Self Portrait to experiment with different styles and parameters and to explore the possibilities of Neural Style Transfer. This is the selection of those portraits that are preserved. These images are actually the very first images I managed to generate in very high resolution using an improved version of the original algorithm. Most of these images were created in early 2016 using style images that I had used before with two exceptions: 1) The style image that was used to create "Self Portrait with X" could not be recovered and 2) "Self-portrait with Candy" was created separately in late 2016. "candy.jpg" was the filename of a popular style used in Justin Johnson's Fast Neural Style Transfer follow up work and I was curious to see how that it would come out with the original Neural Style Transfer algorithm.

About the artist

Leon Gatys

Leon Gatys is an AI researcher whose work centers on leveraging data-driven technology to improve human lives. He is currently based in Seattle, Washington, where he supports Apple’s Responsible AI team. Previously, he co-founded PreemptiveAI, a company dedicated to developing foundation models for human health and physiology using data collected from smartphones and wearable devices. Before that, he was a Research Scientist and one of the founding members of Apple’s Health AI team, where he focused on advancing machine learning for health-related applications. Gatys’ academic background is equally notable. During his PhD, under the supervision of Prof. Matthias Bethge, he pioneered the development of Neural Style Transfer, a groundbreaking algorithm that became one of the earliest demonstrations of the creative potential of generative AI. This work has had a lasting impact on both AI research and the broader cultural imagination of what artificial intelligence can achieve.