Unified Vector Floorplan Generation via Markup Representation

1. Previous work mainly focuses on diffusion on vertices, but diffusion-based models could not be applied directly to variable length generation. Therefore, previous methods require pre-defined numbers of rooms and doors (HouseDiffusion [Shabani+, CVPR23], Cons2Plan [Hong+, MM24]), or additional networks for edge extraction and room allocation (GSDiff [Hu+, AAAI25]).
2. Such suboptimal representation for floorplan generation makes it difficult for previous methods to generalize to multiple conditions.

Our proposed "Floorplan Markup Language" unifies floorplan generation tasks into a next-token prediction task.

We represent the structural information of floorplans in a markup-based grammer.

Automatic residential floorplan generation has long been a central challenge bridging architecture and computer graphics, aiming to make spatial design more efficient and accessible. While early methods based on constraint satisfaction or combinatorial optimization ensure feasibility, they lack diversity and flexibility. Recent generative models achieve promising results but struggle to generalize across heterogeneous conditional tasks, such as generation from site boundaries, room adjacency graphs, or partial layouts, due to their suboptimal representations.

To address this gap, we introduce Floorplan Markup Language (FML), a general representation that encodes floorplan information within a single structured grammar, which casts the entire floorplan generation problem into a next token prediction task. Leveraging FML, we develop a transformer-based generative model, FMLM, capable of producing highfidelity and functional floorplans under diverse conditions.

Comprehensive experiments on the RPLAN dataset demonstrate that FMLM, despite being a single model, surpasses the previous task-specific state-of-the-art methods.