The Erasmus MC is getting a digital twin. Scientists of the DE-RISC project use new AI technology to convert 2D floor plans of the building into a 3D model. Interim results from the project were shared during the first meet-up of DE-RISC consortium last month.

The DE-RISC consortium includes a wide variety of stakeholders, varying from scientists from the social and technological field to partners from the public and private sector such as the safety-region Rotterdam-Rijnmond and fire-safety companies. They all gathered at the Convergence Square in the Erasmus MC to discuss their shared ambition: develop a digital infrastructure that can communicate real-time with rescue services to enable for quicker and better evacuation of the building in case of an emergency.

One of the speakers during the event was research student Gijs de Jong from TU Delft. He designed a neural network to synthesize 3D models of buildings form 2D floor plans. ‘Contrary to what I expected, using an effective loss function is just as, or even more, important compared to the architecture of a model. The loss function represents the objective during learning and can be much better tuned to the context of your problem.’ This means that not only the construction of the AI model itself is important, but also how well it can work with different types of data.  The findings from De Jong will facilitate the translation of 2D floor plans to 3D models.

The second speaker at the event was Sabien Dekkers from the Erasmus University. While De Jong focuses on the technical aspects, Dekkers contemplates the societal impact. She examines citizens’ attitudes towards data sharing. Dekkers: ‘Effective functioning of the DE-RISC AI technology requires a lot of data but also the consent of society. The data is used for the common good, but must also protect people’s individual wishes and privacy. In my research, I focus on the perspective and opinion of citizens on this subject.’

These two lines of research illustrate the transdisciplinary nature of this project, integrating different disciplines and interests. The complex challenges that stand in the way of creating a digital twin requires collaboration among people across the entire society. Together they aim to build a robust data platform for digital twins, transform 2D floorplans into 3D models and create the right conditions for data-sharing and stakeholder alignment.

Afterwards, the project aims to scale the technology both by deepening the technology development as by broadening and scaling the scope of the project.