CRI: CI-New: A Community Benchmarking Infrastructure for Birectional Reflectance Distribution Functions

点击翻译按钮获取中文摘要

This project will support the development of a testbed application suite for evaluating material appearance models that describe how light is reflected from surfaces in computer graphics research. Photorealistic image synthesis in computer graphics requires accurate simulation of light transport through a virtual scene starting from the light sources, reflecting off surfaces (based on the material models), before ending at the camera. Hence, material models are a fundamental component in computer graphics systems. The testbed will consist of three integral components: (1) reference implementations of existing material models, (2) a library of material model parameters optimized to best mimic the appearance of real materials, as well as tools to compute new sets of parameters to mimic newly measured materials, and (3) online and offline analysis tools to aid researchers and practitioners to select appropriate models and to aid in the development of more accurate material models. A key feature of the testbed is that when people submit new material model implementations, or new strategies for mimicking real materials, or new tools for assessing the accuracy of models, the testbed will automatically evaluate them against all models and data in the testbed. This key feature is critical as there currently does not exist a good way for researchers to exhaustively compare their work against others, which slows down progress in both research and graphics applications; the testbed will make it much easier to compare new ideas to old ones, as well as to understand which models work best in which practical situations, making appearance modeling more portable and accessible to researchers and practitioners in and outside computer graphics. This testbed will advance research in areas related to computer graphics, as well as directly benefit society through supporting improved fidelity and visual realism in virtual reality applications, product pre-visualization, e-commerce, safety and training simulators, etc.The testbed will be in the form of a repository and a web interface. The repository will be populated by the lead investigator with an initial set of reference implementations of bidirectional reflectance distribution functions (BRDFs), BRDF fitting strategies, and quality assessment metrics. The research community will then be able to contribute new advances to each of these components to the repository. In addition, fitted BRDF parameters that mimic the reflectance behavior of physical materials for all BRDF models, computed with each of the different BRDF fitting strategies available in the code repository will be made publicly available, both on the repository as well as through the web interface. These fitted BRDF parameters will be dynamically updated when new algorithms or improvements are recorded in the repository. The web interface will allow experts and practitioners to explore the fitted BRDF parameters, corresponding visualization, and quality assessments to drive their research or product development.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.