Research map / 3D vision / 3D generation
Controllable 3D Outdoor Scene Generation via Scene Graphs
Yuheng Liu, Xinke Li, Yuning Zhang, Lu Qi, Xin Li, Wenping Wang, Chongshou Li, Xueting Li, Ming-Hsuan Yang
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 edge reconstruction and node classification auxiliary tasks to the GNN (ablation) | scene-graph diffusion · CarlaSG | Jaccard: 0.81 -> 0.93 | Improved |
| add edge reconstruction and node classification auxiliary tasks to the GNN (ablation) | scene-graph diffusion · CarlaSG | MAE: 0.79 -> 0.63 | Improved |
| add node classification on top of edge reconstruction (ablation) | scene-graph diffusion · CarlaSG | MAE: 0.89 -> 0.63 | Improved |
| add edge reconstruction on top of node classification (ablation) | scene-graph diffusion · CarlaSG | Jaccard: 0.83 -> 0.93 | Improved |
| jointly train diffusion and GNN from scratch with post-trained LOC instead of end-to-end from scratch (ablation) | scene-graph diffusion · CarlaSG | MAE: 1.01 -> 0.63 | Improved |
| jointly train diffusion and GNN from scratch with post-trained LOC instead of pre-training all and freezing GNN and LOC (ablation) | scene-graph diffusion · CarlaSG | MAE: 0.79 -> 0.63 | Improved |
| jointly train diffusion and GNN instead of freezing a pre-trained GNN and LOC (ablation) | scene-graph diffusion · CarlaSG | Jaccard: 0.90 -> 0.93 | Improved |
| raise the unconditional proportion in diffusion training beyond 0.1 (ablation) | scene-graph diffusion · CarlaSG | mIoU, MA, Jaccard, MAE | Mixed |
| condition on scene graph with allocation module instead of SG2Im BEM | scene-graph diffusion · CarlaSG | MAE: 0.97 -> 0.63 | Improved |
| condition on scene graph with allocation module instead of LLM text embedding | scene-graph diffusion · CarlaSG | MAE: 1.44 -> 0.63 | Improved |
| condition on scene graph with allocation module instead of LLM text embedding | scene-graph diffusion · CarlaSG | Jaccard: 0.70 -> 0.93 | Improved |
| add scene graph conditioning to unconditional generation | scene-graph diffusion · CarlaSG | mIoU: 68.21 -> 68.69 | No clear effect |
Closest papers by meaning
- SemCity: Semantic Scene Generation with Triplane Diffusion (CVPR 2024 · similarity 0.88)
- R3CD: Scene Graph to Image Generation with Relation-Aware Compositional Contrastive Control Diffusion (AAAI 2024 · similarity 0.87)
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (NeurIPS 2025 · similarity 0.87)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.87)
- MMGDreamer: Mixed-Modality Graph for Geometry-Controllable 3D Indoor Scene Generation (AAAI 2025 · similarity 0.87)
- FreeScene: Mixed Graph Diffusion for 3D Scene Synthesis from Free Prompts (CVPR 2025 · similarity 0.86)
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.
