Visual Computing set the scene for a successful IC Reseach Day

© Alain Herzog / EPFL 2017

© Alain Herzog / EPFL 2017

Over 470 participants took part in the 21th edition of the IC Research Day at the SwissTech Convention Center last Thursday. Three excellent keynote speakers provided an overview of the field of Visual Computing to a captivated audience of researchers, current and past students, professors, staff and representatives of the industry.

To the frontiers of visual computing and back, is where the three keynote speakers’ of the 2017 edition of the IC Research Day took the audience during the one-day event chaired by Professors Mark Pauly and Wenzel Jakob. The conference began with a welcome address of Professor Andreas Mortensen, Vice President for Research, followed by IC Dean James Larus’s up-date on the school’s recent activities.

Dr. Sylvain Paris, of Adobe Research, opened the day demonstrating how to facilitate photo retouching by democratising the tools available, to allow even the layman to reach the levels of results previously only accessible to professional photographers.The researcher proposed the use of algorithms, which transform images based on statistics of superior quality photos as well as machine learning techniques, which would transfer the style of a given photo on another, while preserving the image structure.

IC School’s own, Professor Wenzel Jakob, demonstrated how physics-based image rendering algorithms function, and how tracing light paths; simulating the interaction of light and matter, could be used to produce extraordinarily realistic synthetic images.These techniques are regularly used in film development by the film industry for their exceptional image rendering quality.

Professor Maneesh Agrawala of the University of Stanford presented his recent projects aiming to reduce the effort required to edit and produce high-quality, compelling audio and visual stories. He demonstrated storytelling interfaces, which use representations matching the users’ conceptual models, making it easier to express the overall story design, all the while decreasing the interface complexity significantly.

The storytelling tools shown demonstrated the necessity of design principles, which leverage the knowledge of experts, allowing users to concentrate on the overall shape of a given story and leaving the computer to focus on the low-level details of managing and composing the media.

The event concluded successfully in the awarding of the best demonstrations developed by the school’s researchers. The panel, consisting of the guest lecturers and Dean James Larus, selected Ludovic Barman (LCA1) as the second runner-up with “PriFi - Low-Latency, Tracking-Resistant Communications”, Rémi Lebret (LSIR) as the first runner-up with “Powering Online Children Safety with AI” and Benoît Seguin (DHLAB) as the winner of the Best Demonstration Award with “Visual Search in Digitized Art Collections.”