If you want to collaborate with us, register at seoulai.herokuapp.com and join our slack at seoulai.slack.com. If you want to give a talk fill in this form. To receive (by email) the latest information about upcoming meetups, hackathons and other events organized by Seoul AI, subscribe through this form.
Saturday, May 26, 2018
02:00 AM to 04:00 PM
02:00 - 02:15 Networking
02:15 - 03:00 Configuring neural style transfer for artistic visual variations (Chaehan So)
03:00 - 03:40 Brain Tumor Segmentation (Lars Sjösund)
03:40 - 04:00 Art Science and technology; Artwork using AI (Elly Yaeli Cho)
Chaehan So is a professor of design & marketing research at Hongik University, International School of Design for Advanced Studies (IDAS) in Seoul, South Korea. His main research interests are on experience design, design thinking methodology and applied AI. Together, Dr. So is merging these methods into design psychology - new and psychologically-grounded ways of creating user experience. Dr. So has a strong interdisciplinary intent that arises from his diverse educational background - he has master’s degrees in business (Ecole Supérieure de Commerce de Paris, France) and electrical engineering (Technical University Berlin, Germany) and a Ph.D. in Psychology (Humboldt University of Berlin, Germany). Previously, Dr. So had gained over 13 years consulting experience in agile methodologies and management of online companies.
A well-documented neural style transfer algorithm (Johnson 2016) is explored in four experiments on parameters with direct impact on visual output: number of iterations, learning rate, total variation strength, and content to style weight. Results of these experiments allow to give recommendations for the range of parameter configurations that show satisfactory variations of artistic style.
Lars is working as an AI research engineer at the Sweden based startup Peltarion, and in his spare time he’s running Stockholm AI, the biggest AI community in Sweden. He was an exchange student at KAIST many years ago, and now he spends around 30% of his time in Korea.
We’ll start by going through and get intuition about the U-Net, a common baseline for image segmentation. The specific field of application will be brain tumor segmentation. After this we look at Tiramisu, an extension of U-Net that is much more parameter efficient while achieving SOTA on several benchmarks.
Experienced Artist with a demonstrated history of working in the fine art industry. Skilled in Collage, media art, Contemporary Art, Printmaking, and Art Education. Strong arts and design professional with a Graduate of Slade School of Fine Art and graduate of Art Education from Columbia University NY.
I’d like to present my ideas using AI for my artwork. My works are deeply connected to history and the people’s perception of nature. Recently I have been interested in how machine and human respond to each other and I have been showing a performance piece about it last year. I’d like to continue with this subject using AI as a performer.
Register at meetup.com
Mini AI Hackathon (project presentations)