GSDiff: Synthesizing Vector Floorplans via Geometry-enhanced Structural Graph Generation

TL;DR

GSDiff reframes vector floorplan design as structural graph generation, decoupling node and edge synthesis to directly produce vector floorplans. It combines a diffusion-based node generator with a Transformer-based edge predictor, augmented by a node alignment loss and an edge perception strategy to ensure geometric plausibility and topological correctness. The approach supports unconstrained generation as well as boundary and topology-constrained generation, and experiments show superior performance over state-of-the-art methods on multiple metrics and datasets. This work enables diverse, constraint-aware architectural design outputs with improved geometric consistency and practical usability for automated floorplan synthesis.

Abstract

Automating architectural floorplan design is vital for housing and interior design, offering a faster, cost-effective alternative to manual sketches by architects. However, existing methods, including rule-based and learning-based approaches, face challenges in design complexity and constrained generation with extensive post-processing, and tend to obvious geometric inconsistencies such as misalignment, overlap, and gaps. In this work, we propose a novel generative framework for vector floorplan design via structural graph generation, called GSDiff, focusing on wall junction generation and wall segment prediction to capture both geometric and semantic aspects of structural graphs. To improve the geometric rationality of generated structural graphs, we propose two innovative geometry enhancement methods. In wall junction generation, we propose a novel alignment loss function to improve geometric consistency. In wall segment prediction, we propose a random self-supervision method to enhance the model's perception of the overall geometric structure, thereby promoting the generation of reasonable geometric structures. Employing the diffusion model and the Transformer model, as well as the geometry enhancement strategies, our framework can generate wall junctions, wall segments and room polygons with structural and semantic information, resulting in structural graphs that accurately represent floorplans. Extensive experiments show that the proposed method surpasses existing techniques, enabling free generation and constrained generation, marking a shift towards structure generation in architectural design. Code and data are available at https://github.com/SizheHu/GSDiff.

Figures (8)

• Figure 1: Overview. We represent the floorplan as a structural graph (a) and transform the vector floorplan synthesis into structural graph generation (b). We first generate graph nodes using a diffusion model, then predict the existence of edges between each pair of nodes, and finally extract rooms represented by polygons with semantic labels, resulting in vector floorplans.
• Figure 2: Network architecture of GSDiff. we propose to decouple the structure graph generation into two stages: node generation and edge prediction, which results in the complete structural graph. Finally, all minimal polygonal loops of the structural graph are extracted as rooms to obtain the final vector floorplan.
• Figure 3: Constrained generation. Embeddings for constraints are obtained through respective encoders and used as inputs for node generation and edge prediction. Topology constraints use a Transformer-based encoder, while boundary constraints use a CNN-based encoder.
• Figure 4: A gallery of vector floorplans generated using our framework. From top to bottom: unconstrained generation, topology-constrained generation, boundary-constrained generation, and unconstrained generation with slanted walls.
• Figure 5: Comparison with the ground truth, Graph2Plan, and WallPlan on the boundary-constrained generation. Our method can produce more reasonable floorplans.