𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰 𝗙𝗹𝗼𝗼𝗿𝗽𝗹𝗮𝗻 𝗦𝗲𝗴𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝗨𝘀𝗶𝗻𝗴 𝗦𝗲𝗹𝗳-𝗰𝗼𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗻𝗴 𝗚𝗿𝗮𝗽𝗵 𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝘀 Detailed floorplans of buildings are needed for many applications in civil engineering, such as building information modelling, facility management and construction. In an article recently published in the journal Automation in Construction, we propose an approach based on self-constructing graph neural networks to generate digital floorplans with semantic information from scans of floorplans that are given in analog form. Our experiments show that our approach outperforms previous methods with respect to the correct recognition of room types and structures. https://lnkd.in/eGtBa_f5 #SemanticSegmentation #DeepLearning #AI #BIM #FloorPlans #Research