Research map / 3D vision / 3D generation
Statistical Confidence Rescoring for Robust 3D Scene Graph Generation from Multi-View Images
Qi Xun Yeo, Yanyan Li, Gim Hee Lee
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 confidence rescoring (CR) to JointSSG (ablation) | JointSSG · 3RScan (20 obj, 8 pred) | Rel recall: 32.9 -> 37.9 | Improved |
| add confidence rescoring (CR) to JointSSG (ablation) | JointSSG · 3RScan (20 obj, 8 pred) | Obj mRecall: 45.4 -> 54.1 | Improved |
| add masked feature initialization with DINOv2 on top of CR (ablation) | JointSSG + CR · 3RScan (20 obj, 8 pred) | Rel recall: 37.9 -> 41.1 | Improved |
| add masked feature initialization with DINOv2 on top of CR (ablation) | JointSSG + CR · 3RScan (20 obj, 8 pred) | Obj mRecall: 54.1 -> 61.3 | Improved |
| add residual spatial neighbor GNN (ablation) | Ours without neighbor module · 3RScan (20 obj, 8 pred) | Pred mRecall: 35.8 -> 39.2 | Improved |
| add residual spatial neighbor GNN (ablation) | Ours without neighbor module · 3RScan (20 obj, 8 pred) | Rel recall: 41.1 -> 40.5 | No clear effect |
| use CR during inference (ablation) | Ours · 3RScan (20 obj, 8 pred) | Pred mRecall: 35.7 -> 39.2 | Improved |
| use CR during inference (ablation) | Ours · 3RScan (20 obj, 8 pred) | Rel recall: 41.1 -> 40.5 | Hurt |
| use adaptive confidence alpha instead of fixed 0.8 (ablation) | Ours · 3RScan (20 obj, 8 pred) | Rel recall: 37.8 -> 40.5 | Improved |
| use adaptive confidence alpha instead of fixed 0.2 (ablation) | Ours · 3RScan (20 obj, 8 pred) | Rel recall: 31.0 -> 40.5 | Improved |
| use DINOv2 instead of ResNet18 image features (ablation) | Ours · 3RScan (20 obj, 8 pred) | Rel recall: 38.6 -> 40.5 | Improved |
| use our method with ResNet18 features instead of JointSSG with DINOv2 | Ours dagger · 3RScan (20 obj, 8 pred) | Rel recall: 34.3 -> 38.6 | Improved |
Closest papers by meaning
- Incremental 3D Semantic Scene Graph Prediction From RGB Sequences (CVPR 2023 · similarity 0.91)
- DepWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors (ECCV 2026 · similarity 0.90)
- Learning 3D Semantic Scene Graphs From 3D Indoor Reconstructions (CVPR 2020 · similarity 0.89)
- Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · similarity 0.89)
- Multiview Scene Graph (NeurIPS 2024 · similarity 0.88)
- Towards Unbiased and Robust Spatio-Temporal Scene Graph Generation and Anticipation (CVPR 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.
