ECCV 2016 Daily - Thursday

3D Morphable Eye Region Model for Gaze Estimation 8 ECCV Daily : Thursday Presentations ECCV Daily: Errol Wood, what is your position? Errol Wood: I am a PhD student at the University of Cambridge . ECCV Daily: What is your work about? Errol: Our work is about a new morphable model for the eye region. It’s different from previous ones in that it’s more detailed in the eye, and we use it for gaze estimation by fitting it to an eye image. The main novelty comes in the new model, which has multiple parts and is more detailed than before. It’s very difficult to construct 3D, multiple models of the eye region because scanning the eye is very challenging. We can’t model it using linear models such as BCA for the shape and texture. We need to have a separate eyeball model that we incorporate with an eye region model to lead to a multi-part, 3D model of the eye region. We also have a model for iris color variation as well. That’s quite important in capturing dark eyes and light eyes. This is another novelty we have. ECCV Daily: How did you solve these challenges? Which kind of computer vision techniques helped you to solve them? Errol: It’s a mix of computer vision and graphics. We include an anatomy-based eyeball model and computer animation techniques, posing the eye and making realistic eyeball movements which we can then render in a mixture of computer vision and graphics to match an input image with our rendered eyes. “ We include an anatomy-based eyeball model and computer animation techniques ”