Research map / 3D vision / Face and shape
Building-GAN: Graph-Conditioned Architectural Volumetric Design Generation
Kai-Hung Chang, Chin-Yi Cheng, Jieliang Luo, Shingo Murata, Mehdi Nourbakhsh, Yoshito Tsuji
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 story discriminator to building discriminator (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 1.0464 -> 0.0845 | Improved |
| add building discriminator to story discriminator (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 6.8061 -> 0.0845 | Improved |
| add building discriminator to story discriminator (ablation) | Building-GAN · Synthetic building volumetric design dataset | TPR accuracy: 0.207 -> 0.772 | Improved |
| add story discriminator to building discriminator (ablation) | Building-GAN · Synthetic building volumetric design dataset | Connectivity accuracy: 0.459 -> 0.569 | Improved |
| add positional encoding of story indices (vs raw indices) (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 0.1512 -> 0.0845 | Improved |
| add relative position features on top of PE (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 0.8333 -> 0.1512 | Improved |
| increase voxel GNN layers from 4 to 12 (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 1.0463 -> 0.0845 | Improved |
| increase voxel GNN layers from 4 to 12 (ablation) | Building-GAN · Synthetic building volumetric design dataset | Connectivity accuracy: 0.432 -> 0.569 | Improved |
| apply pointer module every 2 steps instead of only first + last (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 3.3818 -> 0.0845 | Improved |
| apply pointer module every 2 steps instead of every 6 steps (ablation) | Building-GAN · Synthetic building volumetric design dataset | FID: 0.2179 -> 0.0845 | Improved |
| apply pointer module every 2 steps instead of every 6 steps (ablation) | Building-GAN · Synthetic building volumetric design dataset | Connectivity accuracy: 0.547 -> 0.569 | Improved |
| use Building-GAN voxel graph model instead of House-GAN | Building-GAN · Synthetic building volumetric design dataset | FID: 17.6003 -> 0.0845 | Improved |
Closest papers by meaning
- House-GAN++: Generative Adversarial Layout Refinement Network towards Intelligent Computational Agent for Professional Architects (CVPR 2021 · similarity 0.89)
- Intelligent Home 3D: Automatic 3D-House Design From Linguistic Descriptions Only (CVPR 2020 · similarity 0.88)
- House-GAN: Relational Generative Adversarial Networks for Graph-constrained House Layout Generation (ECCV 2020 · similarity 0.87)
- Graph Transformer GANs With Graph Masked Modeling for Architectural Layout Generation (TPAMI 2024 · similarity 0.87)
- GSDiff: Synthesizing Vector Floorplans via Geometry-enhanced Structural Graph Generation (AAAI 2025 · similarity 0.86)
- Revisiting CAD Model Generation by Learning Raster Sketch (AAAI 2025 · similarity 0.84)
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.
