Research map / Image generation / GANs
House-GAN: Relational Generative Adversarial Networks for Graph-constrained House Layout Generation
Nelson Nauata, Kai-Hung Chang, Chin-Yi Cheng, Greg Mori, Yasutaka Furukawa
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add room connectivity to room count and type (ablation) | House-GAN · House layout, 13+ rooms | compatibility graph edit distance: 21.4 -> 10.8 | Improved |
| add room connectivity to room count and type (ablation) | House-GAN · House layout, 4-6 rooms | compatibility graph edit distance: 2.2 -> 1.1 | Improved |
| add room type to room count (ablation) | House-GAN · House layout, 13+ rooms | compatibility graph edit distance: 37.3 -> 21.4 | Improved |
| add room count (relational graph) to no graph information (ablation) | House-GAN · House layout, 13+ rooms | compatibility graph edit distance: 32.2 -> 37.3 | Hurt |
| add room count (relational graph) to no graph information (ablation) | House-GAN · House layout, 1-3 rooms | compatibility graph edit distance: 28.5 -> 0.6 | Improved |
| replace GCN vector message passing with Conv-MPN feature volumes | House-GAN · House layout, 13+ rooms | diversity FID: 31.5 -> 20.1 | Improved |
| replace GCN with Conv-MPN | House-GAN · House layout, 13+ rooms | compatibility graph edit distance: 3.7 -> 10.8 | Hurt |
| replace CNN-only generator with relational House-GAN | House-GAN · House layout, 13+ rooms | diversity FID: 90.0 -> 20.1 | Improved |
| replace GCN with Conv-MPN | House-GAN · House layout, all groups | realism user score: 0.11 -> 0.15 | Improved |
| add spectral normalization or per-room discriminator (ablation) | House-GAN · House layout | quality | No clear effect |
Later papers that compared with this method
5 later accepted papers report HouseGAN++ as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- CG-Floor: Centroid-Guided Diffusion for Large-Scale Floorplan Generation (CVPR 2026 · better on 3 of 3 shared results)
- Unified Vector Floorplan Generation via Markup Representation (CVPR 2026 · better on 2 of 2 shared results)
- HouseDiffusion: Vector Floorplan Generation via a Diffusion Model With Discrete and Continuous Denoising (CVPR 2023 · better on 11 of 11 shared results)
- House-GAN++: Generative Adversarial Layout Refinement Network towards Intelligent Computational Agent for Professional Architects (CVPR 2021 · better on 10 of 10 shared results)
- Roof-GAN: Learning To Generate Roof Geometry and Relations for Residential Houses (CVPR 2021 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- House-GAN++: Generative Adversarial Layout Refinement Network towards Intelligent Computational Agent for Professional Architects (CVPR 2021 · similarity 0.93)
- Graph Transformer GANs for Graph-Constrained House Generation (CVPR 2023 · similarity 0.92)
- Graph Transformer GANs With Graph Masked Modeling for Architectural Layout Generation (TPAMI 2024 · similarity 0.90)
- Roof-GAN: Learning To Generate Roof Geometry and Relations for Residential Houses (CVPR 2021 · similarity 0.90)
- Constrained Layout Generation with Factor Graphs (CVPR 2024 · similarity 0.85)
- GSDiff: Synthesizing Vector Floorplans via Geometry-enhanced Structural Graph Generation (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
