Research map / 3D vision / 3D generation
Hierarchical 3D Scene Graphs Construction Outdoors
Jon Nyffeler, Federico Tombari, Daniel Barath
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 structure (S) modality to point cloud (P) training (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on LIN | mean reciprocal rank: 0.963 -> 0.939 | Hurt |
| add structure (S) modality to point cloud (P) training (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on CAB | mean reciprocal rank: 0.944 -> 0.865 | Hurt |
| add structure (S) modality to point cloud (P) training (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on HGE | mean reciprocal rank: 0.942 -> 0.892 | Hurt |
| add hyperedge (H) modality to point cloud (P) training (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on LIN | mean reciprocal rank: 0.963 -> 0.967 | Improved |
| add hyperedge (H) modality to point cloud (P) training (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on CAB | mean reciprocal rank: 0.944 -> 0.938 | Hurt |
| add hyperedge (H) modality to point cloud (P) training (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on HGE | mean reciprocal rank: 0.942 -> 0.944 | Improved |
| use hyperedge plus attribute (H + A) modalities instead of point cloud only (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on LIN | mean reciprocal rank: 0.963 -> 0.996 | Improved |
| use hyperedge plus attribute (H + A) modalities instead of point cloud only (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on HGE | mean reciprocal rank: 0.942 -> 0.971 | Improved |
| use hyperedge plus attribute (H + A) modalities instead of point cloud only (ablation) | SGAligner on our scene graphs · LaMAR node matching, trained on CAB | mean reciprocal rank: 0.944 -> 0.941 | No clear effect |
| replace direct FPFH registration with SGAligner object matching on our scene graphs (FPFH) | FPFH registration · LaMAR LIN point cloud registration | RRE (degrees): 32.049 -> 0.279 | Improved |
| replace direct FPFH registration with SGAligner object matching on our scene graphs (FPFH) | FPFH registration · LaMAR CAB point cloud registration | RTE (meters): 14.453 -> 1.552 | Improved |
| replace direct FPFH registration with SGAligner object matching on our scene graphs (FPFH) | FPFH registration · LaMAR HGE point cloud registration | runtime (s): 2.644 -> 5.981 | Hurt |
Closest papers by meaning
- Open3DSG: Open-Vocabulary 3D Scene Graphs from Point Clouds with Queryable Objects and Open-Set Relationships (CVPR 2024 · similarity 0.88)
- USCILab3D: A Large-scale, Long-term, Semantically Annotated Outdoor Dataset (NeurIPS 2024 · similarity 0.87)
- SemCity: Semantic Scene Generation with Triplane Diffusion (CVPR 2024 · similarity 0.87)
- TB-HSU: Hierarchical 3D Scene Understanding with Contextual Affordances (AAAI 2025 · similarity 0.87)
- Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · similarity 0.86)
- Open-Vocabulary Functional 3D Scene Graphs for Real-World Indoor Spaces (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.
