Research map / 3D vision / 3D generation
Pair2Scene: Learning Local Object Relations for Procedural Scene Generation
Xingjian Ran, Shujie Zhang, Weipeng Zhong, Luo Li, Bo Dai
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 |
|---|---|---|---|
| model local object relations instead of global coordinate regression (ablation) | Pair2Scene · 3D-Front only | FID: 92.34 -> 65.15 | Improved |
| pretrain Point-MAE geometric encoder on the asset library (ablation) | Pair2Scene · 3D-Front only | FID: 81.14 -> 65.15 | Improved |
| add collision-aware rejection sampling (ablation) | Pair2Scene · 3D-Front only | FID: 78.26 -> 65.15 | Improved |
| use point cloud geometry instead of class-only embeddings (ablation) | Pair2Scene · 3D-Front only | FID: 68.45 -> 65.15 | Improved |
| add gravity simulation after placement (ablation) | Pair2Scene · 3D-Front only | FID: 66.82 -> 65.15 | No clear effect |
| let Pair2Scene expand beyond dataset object counts (Ours-Beyond vs Ours-Fit) | Pair2Scene · 3D-Front only | FID: 65.92 -> 75.88 | Hurt |
| let Pair2Scene expand beyond dataset object counts (Ours-Beyond vs Ours-Fit) | Pair2Scene · 3D-Front only user study | Contextual Fidelity Score: 2.73 -> 4.46 | Improved |
| use Pair2Scene local-rule model instead of global layout predictors | ATISS and DiffuScene · 3D-Front only | KID (x1e-3): 31.72 -> 22.14 (vs DiffuScene) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DiffuScene — DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis (CVPR 2024) · better on 5 of 5 shared results
- Holodeck — Holodeck: Language Guided Generation of 3D Embodied AI Environments (CVPR 2024) · better on 2 of 2 shared results
- ATISS — ATISS: Autoregressive Transformers for Indoor Scene Synthesis (NeurIPS 2021) · better on 2 of 2 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
- InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Prior (ICLR 2024 · similarity 0.89)
- Taming LLMs for Codematic Indoor Scene Generation (ECCV 2026 · similarity 0.89)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.89)
- From Programs to Poses: Factored Real-World Scene Generation via Learned Program Libraries (NeurIPS 2025 · similarity 0.88)
- SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (NeurIPS 2025 · similarity 0.88)
- InternScenes: A Large-scale Simulatable Indoor Scene Dataset with Realistic Layouts (NeurIPS 2025 · similarity 0.87)
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.
