The objectives of this work are the design and creation of a graph-based representation of a building, suitable for training Deep Learning (DL) methods. The informational background for the graph-based representation is a building model created with Building Information Modeling (BIM). In this context, the research question is raised: How can the graph-based representation of a BIM model be developed and created to support training data for Deep Learning methods? In order to answer this research question, an initial investigation was started to determine which information should be selected in the building representation to make the structure of the graph-based representation as precise as possible. The result of the work is a method that can be easily applied to existing BIM models and generates a network representation from them. Finally, a visual representation of an example model is explained. Based on the generated network representation training datasets for e.g. DL methods can be created.     «

The objectives of this work are the design and creation of a graph-based representation of a building, suitable for training Deep Learning (DL) methods. The informational background for the graph-based representation is a building model created with Building Information Modeling (BIM). In this context, the research question is raised: How can the graph-based representation of a BIM model be developed and created to support training data for Deep Learning methods? In order to answer this research...     »