SBIR Phase II: Fast Creation of Photorealistic 3D Models using Consumer Hardware

这个小企业创新研究（SBIR）二期项目的广泛影响/商业潜力将是巨大的：一个成功的项目将改变建筑行业，通过减少法律冲突、进度延误和糟糕的决策，使其更有效率。建议的工作将能够快速轻松地创建物理空间的100%完整视觉文档；该文档可以在整个构建过程中多次生成。这样，拟议的项目将使建筑行业的专业人士能够比以往更有效地跟踪进度，并与他们的团队进行沟通。该项目的第二个成果将是创建庞大、详细、从未见过的建筑项目和房地产数据集，使人工智能和计算机视觉方面的技术创新能够影响美国乃至世界上最大的行业之一。例如，可以训练系统自动发现安全问题，增加安全管理人员的努力，使工人比以往任何时候都更安全。这个小型企业创新研究（SBIR）二期项目将开发一种快速、易于使用和廉价的方法，使用现成的消费者硬件创建逼真的沉浸式模型。技术障碍包括验证使用消费者硬件生成的模型的质量和有效性，以及在没有人工注释的情况下通过3D空间自动创建路线。技术里程碑包括使用各种传感器流以及其他先前数据来构建这些路线。随着这些障碍的清除，先进的工作可能包括随着时间推移在同一地点捕获的3D模型之间进行自动分析。由于该系统易于使用，它将能够收集大量全新的数据集。这项研究的目标是产生一个原型，外行人可以用它来创建一个身临其境的物理站点模型，以便在没有注释的情况下记录它。达到这些目标的计划包括针对上面列出的障碍的迭代软件开发，以及持续的用户反馈来指导和改进开发。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase II project will be substantial: a successful project would transform the construction industry, making it far more efficient by reducing legal conflicts, schedule slips and poor decision making. The proposed work will enable the fast and easy creation of 100% complete visual documentation of a physical space; this documentation can be generated many times throughout the course of construction. In so doing, the proposed project will allow professionals in the construction industry to track progress and communicate with their teams far more efficiently than ever before. A second outcome of the project will be the creation of vast, detailed, never before seen datasets of construction projects and real estate, allowing technical innovations in artificial intelligence and computer vision to impact one of the largest industries in the nation and the world. For example, systems could be trained to automatically spot safety concerns, augmenting the efforts of safety managers and keeping workers safer than ever before.This Small Business Innovation Research (SBIR) Phase II project will develop a fast, easy to use and cheap method to create photorealistic immersive models using off the shelf consumer hardware. Technical hurdles include validating the quality and efficacy of models generated with consumer hardware and automatic creation of routes through the 3D space without human annotation. Technical milestones involve using various sensor streams as well as other prior data to build these routes. With these hurdles cleared, advanced work may include automated analytics between and among 3D models of the same site captured over time. Because of the system's ease of use, it will enable the collection of large, totally novel datasets. The goal of the research is to produce a prototype that a layperson can use to create an immersive model of a physical site in order to document it with no annotation effort. The plan to reach these goals includes iterative software development against the hurdles listed above, as well as continuous user feedback to guide and refine 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.