Deep Learning Through Parametrically Generated Virtual Building Information Models for Real-World Object Recognition

The present research applies parametric building information modeling (BIM) to automatically generate big training data for machine learning to recognize building objects in the real world, which would be useful for various applications including surveys, documentation, and performance modeling. Training data is typically manually curated and annotated by people. We trained ANNs solely on realistic images of parametrically generated 3D BIM models using the BIMGenE program developed in this research. The ANN training result demonstrated generalizability and good semantic segmentation on a test case as well as arbitrary photos of buildings, which is significant for the future of training AI with generated data for solving real-world architectural problems.