Research map / 3D vision / 3D generation
CommonScenes: Generating Commonsense 3D Indoor Scenes with Scene Graph Diffusion
Guangyao Zhai, Evin Pinar Örnek, Shun-Cheng Wu, Yan Di, Federico Tombari, Nassir Navab, Benjamin Busam
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 |
|---|---|---|---|
| replace DeepSDF shape decoding with relation-conditioned latent diffusion (rel2shape) | Graph-to-3D style pipeline · SG-FRONT all rooms | FID: 50.29 -> 45.70 | Improved |
| replace DeepSDF shape decoding with relation-conditioned latent diffusion | Graph-to-3D style pipeline · SG-FRONT all rooms | KID (x0.001): 7.96 -> 3.84 | Improved |
| learn layout and shape jointly instead of layout network plus text-to-shape model | CommonScenes · SG-FRONT all rooms | FID: 50.58 -> 45.70 | Improved |
| add contextual graph with CLIP features and box embeddings (compare w/o shape branch to retrieval baselines) | Layout branch · SG-FRONT all rooms | FID: 43.51 -> 42.07 | Improved |
| add diffusion shape branch with retrieval-free generation (ablation) | CommonScenes · SG-FRONT all rooms | FID: 42.07 -> 45.70 | Hurt |
| use rel2shape instead of layout+txt2shape | CommonScenes · SG-FRONT dining room | FID: 64.02 -> 65.71 | Hurt |
| replace Graph-to-3D with CommonScenes | CommonScenes · SG-FRONT generation | symmetrical constraint accuracy: 0.57 -> 0.60 | Improved |
| replace Graph-to-3D with CommonScenes | CommonScenes · SG-FRONT generation | close by constraint accuracy: 0.74 -> 0.77 | Improved |
| replace Graph-to-3D with CommonScenes | CommonScenes · SG-FRONT manipulation | symmetrical accuracy (change mode): 0.46 -> 0.59 | Improved |
| replace Graph-to-3D with CommonScenes | CommonScenes · SG-FRONT diversity | Chamfer Distance x0.001: 1.21 -> 73.40 | Improved |
| remove CLIP contextual features, relation encoder E_r, or use concatenation instead of cross-attention (ablation) | CommonScenes · SG-FRONT | FID/KID | Improved |
| replace Graph-to-3D with CommonScenes on easy relations | CommonScenes · SG-FRONT generation | left/right accuracy: 0.98 -> 0.98 | No clear effect |
Later papers that compared with this method
3 later accepted papers report CommonScenes as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- La La LiDAR: Large-Scale Layout Generation from LiDAR Data (AAAI 2026 · better on 1 of 1 shared result)
- Planner3D: LLM-enhanced Graph Prior Meets 3D Indoor Scene Explicit Regularization (TPAMI 2025 · better on 9 of 10 shared results)
- EchoScene: Indoor Scene Generation via Information Echo over Scene Graph Diffusion (ECCV 2024 · better on 5 of 8 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Graph-to-3D — Graph-to-3D: End-to-End Generation and Manipulation of 3D Scenes Using Scene Graphs (ICCV 2021) · better on 10 of 10 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
- FreeScene: Mixed Graph Diffusion for 3D Scene Synthesis from Free Prompts (CVPR 2025 · similarity 0.90)
- DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis (CVPR 2024 · similarity 0.90)
- InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Prior (ICLR 2024 · similarity 0.90)
- MMGDreamer: Mixed-Modality Graph for Geometry-Controllable 3D Indoor Scene Generation (AAAI 2025 · similarity 0.89)
- SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (NeurIPS 2025 · similarity 0.89)
- SceneCraft: Layout-Guided 3D Scene Generation (NeurIPS 2024 · similarity 0.89)
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.
