3Dfpdesign: a Deep Learning Based 3D Floorplan Drawing Application

This paper presents an optimized approach and development of the 3DFPDesign application for reconstructing 3D floorplans in three-dimensional space, utilizing user-defined boundaries and constraints for residential structures. The application allows users to input house parameters and specify constraints such as room types and quantities. Deep learningbased algorithms are employed to autonomously divide the house boundary, generating a customized and optimized interior layout based on user-defined constraints. The integration of the PfG (Plan From Graph) module stands as the main innovation, transforming user-provided boundary data into structured layout graphs. Drawing inspiration from Graph2Plan [14] model, our system autonomously allocates and refines room divisions, ensuring a harmonious spatial arrangement. This paper proposes a flexible and adaptive solution for personalized home planning within specified boundaries. The application has been thoroughly tested to showcase its features, including the ability to meet userprovided constraints, deliver quick response times, and provide a convenient user interaction experience.