Modeling and representing materials in the wild

Our everyday life brings us in contact with a rich range of materials that contribute to both the utility and aesthetics of our environment. Human beings are very good at using subtle distinctions in appearance to distinguish between materials (e.g., silk vs. cotton, laminate vs. granite). Capturing these visually important, yet subtle, distinctions is critical for applications in many domains: in virtual and augmented reality fueled by the advent of devices like Google Glass, in virtual prototyping for industrial design, in ecommerce and retail, in textile design and prototyping, in interior design and remodeling, and in games and movies. Understanding how humans perceive materials can drive better graphics and vision algorithms for material recognition and understanding, and material reproduction. As a first step towards achieving this goal, it is useful to collect information about the vast range of materials that we encounter in our daily lives. We introduce two new crowdsourced databases of material annotations to drive better material-driven exploration. OpenSurfaces is a rich, labeled database consisting of thousands of examples of surfaces segmented from consumer photographs of interiors, and annotated with material parameters, texture information, and contextual information. IIW (Intrinsic Images in theWild) is a database of pairwise material annotations of points in images that is useful for decomposing images in the wild into material and lighting layers. Together these databases can drive various material-based applications like surface retexturing, intrinsic image decomposition, intelligent material-based image browsing, and material design.