CG-Floor: Centroid-Guided Diffusion for Large-Scale Floorplan Generation

Large-scale floorplan generation is critical for virtual space planning and architectural simulation. Although existing methods have shown success in generating smallscale floorplans with simple room shapes, they struggle to handle complex room connections and irregular room shapes that arise in large-scale floorplans. In this paper, we propose CG-Floor, a centroid-guided hierarchical framework that explicitly decouples room position and shape generation to address these issues. We first introduce the sizeaware semantic centroid heatmap, derived from predicted room centroids and sizes, which provides a structured representation to guide the effective generation of a coarseto-fine floorplan generator while ensuring semantic alignment. Additionally, we train a vector quantized codebook of floorplans with complex room shapes to capture the diversity of room shapes and employ a latent diffusion transformer to generate large-scale floorplans featuring non-Manhattan room shapes. CG-Floor achieves state-of-theart performance on the large-scale MSD dataset, and sup- † Equal contribution.