DigiBygg – AI for automated production of digital building models and smart building analyses.

The DigiBygg project aims to achieve fast and cost-effective production of detailed digital building models, which will promote broader adoption of digital processes in the building sector. To achieve this, new methods based on artificial neural networks will be developed for processing scanned data from buildings to automatically generate floor plans and detailed 3D models. Buildings are the single largest consumer of energy in Europe, responsible for approximately 40% of EU energy consumption and 36% of the energy-related greenhouse gas emissions1. Transformation within the building sector is therefore crucial to achieving the UN sustainability goals. Exact Geo Survey is a leading surveying, scan to BIM, IT and digitisation company that wants to set a new standard for how 3D models and drawings are created from scanned point clouds. This ambitious task will be undertaken together with SINTEF as R&D partner as well as other BIM stakeholders from the entire value chain. The increased availability of scanning technology has greatly simplified the process for collecting data used to construct BIMs. However, creating useful 3D models and plans from a point cloud and image data is still a labour intensive and thus expensive process, presenting a critical barrier to adoption of digital processes, particularly within the municipal and small business sectors. New methods are therefore needed for fast and cost-effective production of 3D models and floor plans from scanned data. In this project, we will build on recent advances to the state of the art for neural network-based processing of 3D data and generative AI to develop novel BIM production technologies that will enable Exact Geo Survey to revolutionise the use of digital processes within the building sector.

This project aims to significantly reduce the real cost and lead time associated with producing digital building models whilst at the same time preserving the accuracy and their level of detail to support the generation of 2D/3D architectural plans and depth of information to support the full range of BIM analyses. We also want to explore how Artificial Intelligence (AI) can be leveraged to revolutionise how digital building models are employed by end-users and to greatly simplify typical analysis workflows – e.g. by enabling natural language queries of building properties and characteristics. We believe a generic and easy to use interface for extracting non-trivial characteristics from digital building models together with greatly reduced cost and lead time will promote adoption of digital workflows and drive digital transformation. Realising these goals will require the following key innovations: • A more effective solution for labelling/segmenting lidar scan data that supports furnished environments and is precise enough to be used as a basis for producing architectural plans. • A new and automated method for producing high-quality 2D plans and 3D models that are in accordance with architectural, construction and private sector requirements. • A new method for automatic classification of building materials. • A new framework for integrating AI models with BIM data to enable natural language queries of building properties and characteristics.