CI-P: Planning a Community Benchmarking Infrastructure for Bidirectional Reflectance Distribution Functions

该项目将支持规划和建立一个初步的试验台应用程序，用于评估材料模型，这些模型描述了计算机图形学研究中光如何从表面反射。在与社区中感兴趣的研究人员交谈后，首席研究员将建立一个应用程序，允许任何人提交和下载进行这项研究的三种重要数据类型：(1)材料模型的参考实现，(2)模型预测的光反射与实际示例的匹配程度的测量，以及(3)物理材料测量反射的示例进行比较。当人们提交新的材料模型实现时，应用程序将使用所有可用的度量和度量示例自动运行它包含的所有模型。这是一个重要的进步，因为目前研究人员还没有很好的方法来将他们的工作与他人的工作进行比较，这减慢了研究和图形应用的进展；测试平台将使比较新想法和旧想法以及理解哪种模型在哪种情况下工作得最好变得更加容易。这将推动与计算机图形学相关的许多领域的研究，并通过改进图形在虚拟现实、动画和数据可视化等实际应用中的应用，直接造福社会。为此，首席研究员将创建一个存储库，其中包含一组初始参考算法、数据集、拟合函数和基于双向反射分布函数（brdf）的现有研究的评估指标。其他研究人员将能够通过应用程序编程接口（API）为存储库做出贡献，该接口允许轻松添加新项目；然后，它将动态地报告与存储库中现有项目进行的全对全比较。在此计划期间，首席研究者将开发初始的基础设施，以演示概念验证，并呈现一个具体的工件，该工件可以启动关于如何使存储库对更广泛的研究社区更有用的讨论。虽然从最直接参与brdf的研究人员开始，但该社区最终将包括来自许多相关研究领域的人员，包括非线性优化、人类感知、渲染、逼真可视化和计算机视觉。

This project will support the planning and building of an initial testbed application for evaluating material models that describe how light is reflected from surfaces in computer graphics research. After talking with interested researchers in the community, the lead investigator will build an application that allows anyone to submit and download three important types of data for doing this research: (1) reference implementations of material models, (2) measures of how well the predicted light reflections of a model match real examples, and (3) examples of measured reflections from physical materials to compare against. When people submit new material model implementations, the application will automatically run all of the models it contains using all of the measures and measured examples available. This is an important advance because currently there is no good way for researchers to 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 situations. This will advance research in a number of areas related to computer graphics, as well as directly benefiting society through improving graphics in real applications including virtual reality, animation, and data visualization. To do this, the lead investigator will create a repository with an initial set of reference algorithms, datasets, fitting functions, and evaluation metrics based on existing research around bidirectional reflectance distribution functions (BRDFs). Other researchers will be able to contribute to the repository through an Application Programming Interface (API) that allows for easy addition of new items; it will then dynamically report on run all-against-all comparisons with existing items in the repository. During this planning grant the lead investigator will develop the initial infrastructure both to demonstrate a proof-of-concept and to present a concrete artifact that can jump start the discussion of how to make the repository more useful to the broader research community. Although starting with researchers most directly involved with BRDFs, the community will eventually include people from a number of related research areas, including non-linear optimization, human perception, rendering, photorealistic visualization, and computer vision.