Theory and Applications of Material Appearance Perception
Staff - Faculty of Informatics
Date: 9 July 2021 / 14:00 - 15:00
Manuel Lagunas, Universidad de Zaragoza, Spain
Observing, recognizing, or comparing materials is a fundamental part of our daily life. Under typical viewing conditions, we are capable of effortlessly identifying the objects that surround us and recognizing the materials they are made of; or we can even imagine the appearance of such objects under different illumination conditions, or if they were present in an object with a different shape. However, understanding the underlying perceptual processes that take place to discern the visual properties of an object, or to create a visual impression of its appearance in our brain, is a long-standing problem yet unsolved. Nowadays, the irruption of deep learning represents a new means of approaching such problem from a different perspective. In this presentation, some of the recent challenges in human visual perception concerning material appearance will be discussed. We will introduce a novel image-based material appearance similarity measure derived directly from a learned perceptual feature space, which correlates with human similarity judgments, we will talk about the challenges that illumination and/or geometry represent to our visual system when recognizing different materials, and last, we will present an application for an image-based full-body human relighting method where material appearance is explicitly modeled.
Manuel is a fourth-year Ph.D. student at Universidad de Zaragoza (Spain) inside the Graphics and Imaging Lab. His Ph.D. thesis is being co-supervised by Diego Gutierrez and Belen Masia, and his main research topics lie in the interface between deep learning, computer graphics, and human perception. During the summer of 2019 and 2020, Manuel was a research intern in Adobe Research, San Jose. He obtained his Bachelor and Master's degrees at Universidad de Zaragoza, majoring in Computer Science and Applied Maths, respectively.
Host: Prof. Piotr Didyk
You can join here