Deep Learning for Automated 3D Floor Plan Generation

Abstract

Abstract: This paper presents an optimized method for reconstructing 3D floor plans using userdefined boundaries and constraints for residential structures. This approach allows users to provide
architectural constraints such as room types and quantities, as well as manual sketches or existing
images of the house boundaries. Advanced deep learning algorithms are used to automatically
partition the house boundaries and create a customized, optimized interior layout based on the users’
architectural constraints. In the experiment phase, we integrated the Graph2Plan-based deep learning
module, which converts the user-provided boundary data and architectural constraints into a
structured 2D floor plan, automatically allocating and refining the rooms to ensure a harmonious
spatial arrangement. The evaluation of the deep learning model’s performance shows that this is a
useful and time-saving solution for designers. Then, we utilized graphics and image processing
techniques to generate the 3D floor plans. Based on this solution, we have developed a 3D floor plan
generation application that provides a flexible and adaptive solution for individual home planning
within defined boundaries. The application has been thoroughly tested to demonstrate its features,
including the ability to meet users’ architectural constraints, provide rapid response times, and offer
a convenient user interaction experience.
Keywords: 3DFloorplan, Floorplan generation, layout graph, RPLAN dataset, house plan.