Inria@SiliconValley Post-Doc: George Koulieris presents his work

George Koulieris

George Alex Koulieris

George Alex Koulieris received his PhD from the School of Electronic and Computer Engineering of the Technical University of Crete, Greece in Computer Graphics and Perception. George’s PhD work focused on human gaze prediction that is based on semantic features of the rendered scene (inter-objects relationships and general context). The novel gaze prediction algorithms developed were then applied to Game Level Design to adjust game level difficulty, Level-of-Detail rendering to improve rendering performance on mobile devices and Stereo Disparity Manipulation based on fixated objects.
George started collaborating with GraphDeco (formerly REVES) during a 6-month internship in 2008 working on soft shadows for Computer Graphics, then during his PhD studies and finally returned as a post-doc: since October 2015, he is the recipient of an Inria@SiliconValley post-doc grant which allows him to contribute to a research collaboration between the GRAPHDECO Inria project-team and BanksLab in UC Berkeley, USA (CRISP associate team).

Stereo (3D) displays are becoming mainstream, especially with the advent of lightweight, cheap and large field-of-view displays such as the Oculus Rift and the Microsoft Hololens, but also with large screens offering cheap but accurate head tracking. Despite this new trend, there are several fundamental problems with stereo vision which are linked to human stereo perception and the graphics rendering algorithms and displays.

The goal of this current work, a collaborative effort between University of California Berkeley (Prof. Martin Banks) and Inria Sophia-Antipolis (Dr. George Drettakis) is to address three issues by taking an inter-disciplinary approach between Computer Science and Human Vision. The vergence accommodation conflict is one of the main causes of visual discomfort on a stereo display; we propose a novel hardware/software design to provide a cheap but effective solution. Our second proposal addresses poor perception of depth in stereo displays, by proposing a new filtering method. Finally, the last topic deals with the important problem of eye position calibration for tracked stereo displays.

Crisp website:

George’s Website: