Digital 3D Expression for 8 Billion People

设计数字3D形状的工具拥有复杂的功能，并产生奇妙的工件。不幸的是，这些工具只有一小部分人享受，他们不仅拥有创造性的灵感，还拥有深奥的技术技能。这项提议将为地球上数十亿人寻求一种3D数字表达的新方法。拟议的研究将沿着三个互补的轨道进行：3D物理模拟的实时方向，与底层数据结构解耦的几何建模，以及在3D表达工作流中寻找内在价值。大自然毫不费力地创造了丰富的三维运动和复杂的几何形式。现有的计算方法要么侧重于精确再现观察到的行为，要么依赖于仔细设置大量不直观的参数。相反，提出的研究将探讨如何艺术地和有表现力地控制物理模拟与少量用户友好的参数。这种改进的计算现实的影响跨越了沉浸式远程体验（例如，教育）和计算制造（例如，通过先进的制造技术进行消费者定制）。现有的数字3D建模工具难以摆脱低级的、技术性的用户界面。这些工具一直被他们对特定的、经典的表面表现的坚持所束缚。拟议的研究将“越野”，通过直接针对故意非标准形状表示的高级用户交互，重新构想3D几何建模。表示本身将在交互过程中根据上下文或可用资源进行调整。这项工作既依赖于也有助于我们对3D几何背景下人工智能的深度学习技术的理解。计算机辅助设计已经成功地改变了我们周围的物体、我们居住的建筑和我们欣赏的数字图像。然而，这种极端实际的动机导致了少数人痛苦地经历了狭窄的数字设计工作流程。大多数3D建模工具都是出了名的困难，被视为达到目的的手段。在这个建议中，我们考虑了数字3D设计工作流程的内在价值，而不是创建的工件。我们的设想是，通过研究和增加3D创作行为的价值，将数字3D设计转化为广泛的3D表达，让每个人都能享受。

Tools for designing digital 3D shapes boast sophisticated functionality and produce marvelous artifacts. Unfortunately, these tools are only enjoyed by the small fraction of the population who possess not just creative inspiration but also esoteric technical skills. This proposal will pursue a novel approach to 3D digital expression for the billions of people on earth. The proposed research will proceed along three complementary tracks:Live direction of 3D physical simulations, Geometric modeling decoupled from underlying data structures, and Finding intrinsic value in 3D expression workflows. Nature effortlessly creates rich three-dimensional motion and complex geometric forms. Existing computational methods either focus on accurate reproduction of observed behaviors or rely on carefully setting a large number of unintuitive parameters. Instead, the proposed research will investigate how to artistically and expressively control physical simulations with a small number of user-friendly parameters. The impact of this improved computational reality span immersive tele-experiences (e.g., education) and computational fabrication (e.g., consumer customization via advanced manufacturing techniques). Existing digital 3D modeling tools struggle to break away from low-level, technical user interfaces. These tools have been hamstrung by their insistence on particular, classic surface representations. The proposed research will go "off road" by reimagining 3D geometric modeling by directly targeting high-level user interactions on deliberately non-standard shape representations. The representation itself will adapt during the interaction depending on context or available resources. This work both depends on and contributes to our state of understanding of deep learning techniques from artificial intelligence in the context of 3D geometry. Computer-aided design has successfully transformed the objects around us, the buildings we inhabit, and the digital images we enjoy. However, this extremely practical motivation has led to narrow digital design workflows painfully experienced by a minority of the population. Most 3D modeling tools are notoriously difficult and seen as a means to an end. In this proposal, we consider the intrinsic value in digital 3D design workflow as opposed to created artifacts. Our hypothesis is that by studying and increasing the value of the act of 3D creation, we will transform digital 3D design into widespread 3D expression to be enjoyed by everyone.