Abstract:Diverse and realistic floor plan data are essential for the development of useful computer-aided methods in architectural design. Today's large-scale floor plan datasets predominantly feature simple floor plan layouts, typically representing single-apartment dwellings only. To compensate for the mismatch between current datasets and the real world, we develop \textbf{Modified Swiss Dwellings} (MSD) -- the first large-scale floor plan dataset that contains a significant share of layouts of multi-apartment dwellings. MSD features over 5.3K floor plans of medium- to large-scale building complexes, covering over 18.9K distinct apartments. We validate that existing approaches for floor plan generation, while effective in simpler scenarios, cannot yet seamlessly address the challenges posed by MSD. Our benchmark calls for new research in floor plan machine understanding. Code and data are open.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that the existing large - scale floor plan datasets mainly contain simple layouts (usually single - family residences) and cannot fully represent the more complex multi - family building layouts in the real world. Specifically, the existing floor plan generation methods work well when dealing with simple scenarios, but perform poorly when dealing with complex multi - family building floor plans. Therefore, the author has developed a new benchmark dataset named Modified Swiss Dwellings (MSD) to fill this gap and promote research on complex building floor plan generation. ### Main problems and goals of the paper 1. **Limitations of existing datasets**: - The existing large - scale floor plan datasets (such as RPLAN and LIFULL) mainly contain simple single - family residence floor plans. - Most of the floor plans in these datasets are axis - aligned rectangular layouts, lacking irregular - shaped rooms common in the real world. - The datasets lack connection relationships between multi - family buildings and necessary structural elements (such as load - bearing walls). 2. **Challenges in generating complex multi - family building floor plans**: - Multi - family building floor plans not only contain more areas to be arranged, but also the connectivity between apartments is very important. - There are structural constraints (such as stairs, load - bearing walls, etc.), and these structural elements must be kept intact during the design process. - Current methods perform poorly when dealing with these complex layouts, indicating that the existing floor plan generation methods need to be re - evaluated. 3. **Developing a new benchmark dataset**: - The author has developed the MSD dataset, which contains 5,372 annotated floor plan images, covering medium - to - large single - family or multi - family building complexes. - The MSD dataset provides accurate geometric and topological property annotations, including necessary structural components (such as load - bearing walls). 4. **Verifying the performance of existing methods**: - The author uses two baseline methods (diffusion - model - based and segmentation - based methods) to conduct experiments on the MSD dataset and finds that the performance of these methods drops significantly when dealing with complex floor plans. - The experimental results show that existing methods face great challenges when dealing with complex multi - family building floor plans and new research directions and technical improvements are required. ### Summary This paper aims to solve the deficiencies of existing floor plan generation methods in dealing with complex multi - family building layouts by developing the MSD dataset. The MSD dataset provides researchers with a benchmark for complex floor plan generation tasks that is closer to the real world, prompting them to rethink and improve current methods.

MSD: A Benchmark Dataset for Floor Plan Generation of Building Complexes

Floor plan generation: The interplay among data, machine, and designer

DStruct2Design: Data and Benchmarks for Data Structure Driven Generative Floor Plan Design

Adapting HouseDiffusion for conditional Floor Plan generation on Modified Swiss Dwellings dataset

Web3D-based Automatic Furniture Layout System Using Recursive Case-Based Reasoning and Floor Field.

WAFFLE: Multimodal Floorplan Understanding in the Wild

FloorSet -- a VLSI Floorplanning Dataset with Design Constraints of Real-World SoCs

GenFloor: Interactive Generative Space Layout System via Encoded Tree Graphs

End-to-end Generative Floor-plan and Layout with Attributes and Relation Graph

Building3D: An Urban-Scale Dataset and Benchmarks for Learning Roof Structures from Point Clouds

FurniScene: A Large-scale 3D Room Dataset with Intricate Furnishing Scenes

Multi-Unit Floor Plan Recognition and Reconstruction Using Improved Semantic Segmentation of Raster-Wise Floor Plans

Automatic Rendering of Building Floor Plan Images from Textual Descriptions in English

Floor plan graph learning for generative design of residential buildings: a discrete denoising diffusion model

Exploiting 2D Floorplan for Building-scale Panorama RGBD Alignment

Building Floorspace in China: A Dataset and Learning Pipeline

InteriorNet: Mega-scale Multi-sensor Photo-realistic Indoor Scenes Dataset

Intelligent floor plan design of modular high-rise residential building based on graph-constrained generative adversarial networks

Large-Scale Indoor Visual-Geometric Multimodal Dataset and Benchmark for Novel View Synthesis

AUTOMATIC 3D BUILDING MODEL GENERATION USING DEEP LEARNING METHODS BASED ON CITYJSON AND 2D FLOOR PLANS

HouseCrafter: Lifting Floorplans to 3D Scenes with 2D Diffusion Model