Research map / 3D vision / 3D generation
Generative indoor scene synthesis
Generates 3D indoor layouts using scene graphs, diffusion, grammars and language-driven spatial reasoning.
131 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in 3D generation, keyword cluster “layout · indoor · 3d scene”
State History Tried Baselines Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
29 papers in 2023–24 and 62 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 6.4% to 7.0%.
Open questions the data points to
- Papers report DiffuScene on Living room (FID) anywhere from 122.2 to 186.54: comparisons against it depend on whose number you take.
- Papers report ATISS on Living room (FID) anywhere from 117.67 to 176.95: comparisons against it depend on whose number you take.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · DeLay: Robust Spatial Layout Estimation for Cluttered Indoor Scenes (CVPR 2016 · 155 citations)
- 2017 · ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (CVPR 2017 · 4,275 citations)
- 2018 · LayoutNet: Reconstructing the 3D Room Layout From a Single RGB Image (CVPR 2018 · 266 citations)
- 2019 · HorizonNet: Learning Room Layout With 1D Representation and Pano Stretch Data Augmentation (CVPR 2019 · 190 citations)
- 2020 · Total3DUnderstanding: Joint Layout, Object Pose and Mesh Reconstruction for Indoor Scenes From a Single Image (CVPR 2020 · 240 citations)
- 2021 · OpenRooms: An Open Framework for Photorealistic Indoor Scene Datasets (CVPR 2021 · 74 citations)
- 2022 · Vehicle Trajectory Prediction Works, but Not Everywhere (CVPR 2022 · 67 citations)
- 2023 · Diffusion-Based Generation, Optimization, and Planning in 3D Scenes (CVPR 2023 · 184 citations)
- 2024 · DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis (CVPR 2024 · 93 citations)
- 2025 · WonderWorld: Interactive 3D Scene Generation from a Single Image (CVPR 2025 · 36 citations)
- 2026 · FilmSceneDesigner: Chaining Set Design for Procedural Film Scene Generation (AAAI 2026 · 1 citation)
Earliest papers here
- DeLay: Robust Spatial Layout Estimation for Cluttered Indoor Scenes (CVPR 2016 · 155 citations)
- ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (CVPR 2017 · 4,275 citations)
- IM2CAD (CVPR 2017 · 155 citations)
What has been tried
96 papers here report 949 measured changes (a component added, removed or swapped, compared with a baseline): helped 78%, no clear effect 4%, hurt 17%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Replace UNet-1D+Attention with a transformer denoiser | 3D-FRONT bedroom · FID | 17.21 -> 29.08 | Hurt | DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis (CVPR 2024 · 93 citations) |
| Use spatial relational constraints rather than placing objects along room edges | Residential scenes (average) · Mean Reciprocal Rank | 0.645 -> 0.706 | Helped | Holodeck: Language Guided Generation of 3D Embodied AI Environments (CVPR 2024 · 91 citations) |
| use HSSD-122 rather than ProcTHOR-122 | HM3DSem val · success | 9.3 -> 19.2 | Helped | Habitat Synthetic Scenes Dataset (HSSD-200): An Analysis of 3D Scene Scale and Realism Tradeoffs for ObjectGoal Navigation (CVPR 2024 · 55 citations) |
| Add the Building Instance Generator | GoogleEarth evaluation set · FID | 213.56 -> 97.38 | Helped | CityDreamer: Compositional Generative Model of Unbounded 3D Cities (CVPR 2024 · 50 citations) |
| add collision guidance | Living Room · Colobj | 0.200 -> 0.111 | Helped | PhyScene: Physically Interactable 3D Scene Synthesis for Embodied AI (CVPR 2024 · 48 citations) |
| Remove discrete changes | 20 random scenes, 5k solver steps · Avg. Score (s) | 80.84 -> 86.20 | Helped | Infinigen Indoors: Photorealistic Indoor Scenes using Procedural Generation (CVPR 2024 · 43 citations) |
| Add geometry-based initialization and surfel design | Novel view renderings · CS | 27.23 -> 29.47 | Helped | WonderWorld: Interactive 3D Scene Generation from a Single Image (CVPR 2025 · 36 citations) |
| add a 3D semantic proxy room | Room generation · 3D Structure Completeness | 1.78 -> 2.81 | Helped | ControlRoom3D: Room Generation using Semantic Proxy Rooms (CVPR 2024 · 31 citations) |
| remove positional embedding | SemanticKITTI · FID | 56.55 -> 59.14 | Hurt | SemCity: Semantic Scene Generation with Triplane Diffusion (CVPR 2024 · 21 citations) |
| add the 3D Modeling Spatial Localization Mechanism | Image input · IR | 37.15 -> 1.61 | Helped | CAD-GPT: Synthesising CAD Construction Sequence with Spatial Reasoning-Enhanced Multimodal LLMs (AAAI 2025 · 20 citations) |
| Use Task and Document together rather than Document alone | Tree asset generation · ER@1 | 42.00 -> 84.00 | Helped | SceneX: Procedural Controllable Large-Scale Scene Generation (AAAI 2025 · 13 citations) |
| Use a general base prompt instead of BLIP2 captions during finetuning | Layout-guided image generation · Layout-following ability | Helped | SceneCraft: Layout-Guided 3D Scene Generation (NeurIPS 2024 · 12 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| DiffuScene | Living room | FID | 122.2 – 186.54 | 2 papers, e.g. ICLR 2024 and NeurIPS 2025 |
| ATISS | Living room | FID | 117.67 – 176.95 | 2 papers, e.g. ICLR 2024 and NeurIPS 2025 |
| DiffuScene | Bedroom | FID | 123.09 – 135.57 | 2 papers, e.g. ICLR 2024 and NeurIPS 2025 |
What reviewers will expect
Datasets and baselines named in the 35 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- 3D-FRONT · 8 of 35 papers
- 3D-FUTURE · 6 of 35 papers
- Objaverse · 5 of 35 papers
- ScanNet · 4 of 35 papers
- PartNet-Mobility · 2 of 35 papers
- ShapeNet · 2 of 35 papers
- Matterport3D · 2 of 35 papers
- LLFF · 2 of 35 papers
Baselines
- ATISS · 7 of 35 papers
- DiffuScene · 7 of 35 papers
- LayoutGPT · 7 of 35 papers
- Holodeck · 5 of 35 papers
- PhyScene · 3 of 35 papers
- EchoScene · 2 of 35 papers
- Sync2Gen · 2 of 35 papers
- Text2Room · 2 of 35 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 91 papers; from OpenAlex and the papers' affiliations).
- Shanghai Jiao Tong University · 8 papers
- Stanford University · 7 papers
- Google · 6 papers
- Technical University of Munich · 5 papers
- Shanghai AI Laboratory · 4 papers
- Tencent · 3 papers
- Adobe · 3 papers
- Meta · 3 papers
- National University of Singapore · 3 papers
- Cornell University · 3 papers
Use it
Download BibTeX of 37 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Open-vocabulary 3D scene understanding (95 papers)
- Visual-acoustic scene modeling (12 papers)
Elsewhere in the topic
- Learning 3D reconstruction from images (196 papers)
- Diffusion-based novel view synthesis (156 papers)
- Semantic scene completion (153 papers)
- Hand-object interaction generation (116 papers)
- Feed-forward 3D asset generation and editing (116 papers)
- Score distillation for text-to-3D (106 papers)
- 2D-to-3D pretrained representations (94 papers)
- Monocular 3D object reconstruction (94 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).