2d StyleGAN, pix2pix [140]; pix2pix, CycleGAN [141] o3 D-3D to FD-3D StyleGAN [142] (o1 P-3D + s2 Plan) to FD-3D GCN [142]; cGAN, GNN [143] formation can be generated based on the input RGB chan- nel generative AI can also generate refined 3D models from architectural floor plans (o3 D-3D to FD-3D) [142], or based on spatial se- quence matrices and spatial requirements (o1 P-3D + s2 Plan to FD-3D) [143 Floorplangan: Vector res- idential floorplan adversarial generation. Automation in Construction, 142:104470, 2022. [103] Morteza Rahbar, Mohammadjavad Mahdavinejad, Amir HD Markazi, and Mohammadreza

[4]沈立．人工智能建筑设计可能性与局限性的初探［J］．绿色建造与智能建筑，2023,(06):68-70. [5]蔡炜杰．人工智能对建筑设计的影响和应用研究［J］．江西建材，2023,(04):142-143+146. [6]陈韵蕾．新技术视角智慧建筑设计研究——以人工智能为例［J］．互联网周刊，2022,(24):28-31. [7]周祥．人工智能算法在建筑设计中的应用探索［J］．中外建筑，2021