Smart control room for construction industry, powered by 3D AI

建筑业是一个价值12万亿美元的行业，生产率和数字化水平都很低。这导致重大延误和成本超支成为常态，每年浪费5万亿美元。2018年，我们发现缺乏对施工进度和性能的实时可见性是一个关键问题。手工报告过程意味着基于直觉做出的决策，而不是基于可操作的见解和风险/问题进行管理，导致大量的返工、低生产率和复合风险。去年，Contilio和伦敦帝国理工学院（ICL）的合作伙伴关系通过“改造建筑”挑战获得了一项重要的Innovate UK研发资助，以解决这一关键问题。通过应用尖端技术，我们成功开发了一种基于人工智能的软件，用于自动化破碎建筑的进度和性能，从而帮助按时，按预算交付建筑和基础设施项目，并大大提高了生产力和可靠性。我们还对建筑环境的计算机视觉和深度学习方法进行了最先进的研究。在去年成功获得资助的基础上，该项目将进一步推进Contilio的3D深度学习技术，进行最先进的研究，并开发一个互补的软件模块。我们将为建筑环境中的3D深度学习应用开发一种可扩展和自动化的数据标签方法，并创建一个基于人工智能的软件原型，使建筑质量和认证过程完全自动化。我们将比较不同的数据收集技术，以评估是否可以使用低成本的相机输出来收集可靠的高质量现场数据。我们还将研究扫描到bim方法的使用，为创建智能和结构化的数字双胞胎做出重大贡献。该项目由Contilio（一家位于伦敦的快速发展的建筑技术初创公司）和伦敦帝国理工学院计算机系和土木与环境工程系共同开展，并得到了多个行业合作伙伴的支持。它与英国政府建设战略的目标一致，并与整个BIM过程联系在一起。

Construction is a $12Tn industry suffering from very low productivity and digitization. This results in major delays and cost-overruns to be the norm, wasting $5Tn annually. In 2018, we identified the lack of real-time visibility on construction progress and performance to be a keyproblem. The manual reporting process means decisions made based on intuition as opposed to actionable insights and risks/issues managed with significant delay, resulting in major rework, low productivity and compounding risks.Last year, the Contilio and Imperial College London (ICL) partnership was awarded a significant Innovate UK R&D grant through the "transforming construction" challenge to tackle this key problem. Through application of cutting-edge technology, we have successfully developed an AI-based software for automating the broken constructionprogress and performance, thus, helping deliver construction and infrastructure projects on-time, on-budget and with significantly enhanced productivity and reliability. We have also conducted state of the art research into computer vision and deep learning methods for the Built environment.Building on the successful momentum of last year's grant, this project will further advance Contilio's 3D deep learning technology, conduct state-of-the-art research, and develop a complementary software module. We will develop a scalable and automated data labelling approach for 3D deep learning applications in the built environment and create an AI-based software prototype that fully automates the construction quality and certification processes. We will compare different data collection technology to evaluate whether low cost camera outputs could be used to collect reliable quality site data. Wealso will investigate the use of Scan-to-BIM methods, contributing significantly to the creation of smart and structured digital twins.This project is conducted by Contilio, a fast-growing construction tech startup based out of London and Imperial College London's Department of Computing and Department of Civil and Environmental Engineering with the support of multiple industry partners. It is aligned with the objectives of the UK Government Construction Strategy and ties into the overall BIM process.