Generating floorplans for various building functionalities via latent diffusion model

TL;DR

The paper tackles the challenge of generating architectural floorplans across diverse building typologies by proposing a scale-agnostic, two-input latent diffusion framework conditioned on a design brief and a footprint. It employs a two-stage training regime with multi-modal conditioning (text and footprint) and a ControlNet-inspired architecture to fuse inputs, achieving high-fidelity, structurally coherent floorplans for stadiums, offices, apartments, libraries, and auditoriums. Quantitative metrics show strong alignment with real layouts (FID $=22.436$, KID $=1.844$, SSIM $=0.130$, PSNR $=7.596$), while a human-evaluation game demonstrates the generated designs approach the quality of real floorplans ($p=0.001$). The work contributes a new multi-modal floorplan dataset, an open evaluation tool, and a demonstration of scalable, automated design exploration that can augment both professionals and non-experts in architectural planning.

Abstract

In the domain of architectural design, the foundational essence of creativity and human intelligence lies in the mastery of solving floorplans, a skill demanding distinctive expertise and years of experience. Traditionally, the architectural design process of creating floorplans often requires substantial manual labour and architectural expertise. Even when relying on parametric design approaches, the process is limited based on the designer's ability to build a complex set of parameters to iteratively explore design alternatives. As a result, these approaches hinder creativity and limit discovery of an optimal solution. Here, we present a generative latent diffusion model that learns to generate floorplans for various building types based on building footprints and design briefs. The introduced model learns from the complexity of the inter-connections between diverse building types and the mutations of architectural designs. By harnessing the power of latent diffusion models, this research surpasses conventional limitations in the design process. The model's ability to learn from diverse building types means that it cannot only replicate existing designs but also produce entirely new configurations that fuse design elements in unexpected ways. This innovation introduces a new dimension of creativity into architectural design, allowing architects, urban planners and even individuals without specialised expertise to explore uncharted territories of form and function with speed and cost-effectiveness.

Figures (20)

• Figure 1: Generated floorplans illustrating various building types. Our model takes two inputs—lot footprint and textual description—and outputs the corresponding architectural floorplans.
• Figure 2: The overall proposed architecture based on the LDM model.
• Figure 3: Examples of the introduced datasets, showing paired images and their corresponding text prompts.
• Figure 4: Training and validation losses for the introduced model.
• Figure 5: Results of human evaluations for real and generated images in the Floorplan Game. a) Boxplot displaying rating variations for each image. b) Distribution of scores (0-10) for both image types, showing score overlap. c) Boxplot comparing average scores across all players for real and generated images.