Unified Vector Floorplan Generation via Markup Representation

Kaede Shiohara, Toshihiko Yamasaki
The University of Tokyo
CVPR 2026 (Highlight)
arXiv Video Poster Contact

Our "Floorplan Markup Language" (FML) formulates any vector floorplan generation tasks as next-token prediction.

Why is this work important?

Supported floorplan generation tasks
  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.

Core Idea

Core idea of Floorplan Markup Language

We introduce Floorplan Markup Language (FML) that represents the structural information of each floorplan and condition as a single markup-based sentence.

Once converted into FML, vector floorplan generation can be cast into Next-token Prediction.

Core idea of Floorplan Markup Language

Comparison with Previous State-of-the-Arts

Boundary Conditional Generation


Graph Conditional Generation


Boundary-Graph Conditional Generation

More Tasks

Please refer to our research paper.

BibTeX

@InProceedings{Shiohara_2026_CVPR,
    author    = {Shiohara, Kaede and Yamasaki, Toshihiko},
    title     = {Unified Vector Floorplan Generation via Markup Representation},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2026},
    pages     = {39262-39271}
}