Research map / 3D vision / 3D generation
FROSS: Faster-Than-Real-Time Online 3D Semantic Scene Graph Generation from RGB-D Images
Hao-Yu Hou, Chun-Yi Lee, Motoharu Sonogashira, Yasutomo Kawanishi
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 |
|---|---|---|---|
| use ground-truth 2D scene graphs instead of predicted (ablation) | FROSS · 3DSSG | relationship recall: 27.9 -> 55.8 | Improved |
| use ground-truth 2D scene graphs instead of predicted (ablation) | FROSS · 3DSSG | object mRecall: 63.8 -> 93.3 | Improved |
| use ground-truth 2D scene graphs instead of predicted (ablation) | FROSS · ReplicaSSG | relationship recall: 22.3 -> 67.6 | Improved |
| use ground-truth 2D scene graphs instead of predicted (ablation) | FROSS · ReplicaSSG | object recall: 26.1 -> 89.1 | Improved |
| use ORB-SLAM3 estimated trajectories instead of ground-truth poses (ablation) | FROSS · ReplicaSSG | relationship recall: 22.3 -> 22.7 | No clear effect |
| use ORB-SLAM3 estimated trajectories instead of ground-truth poses (ablation) | FROSS · ReplicaSSG | object recall: 26.1 -> 25.8 | No clear effect |
| raise Hellinger merge threshold from 0.65 to 0.85 (ablation) | FROSS · threshold sweep | relationship recall: 16.5 -> 30.0 | Improved |
| raise Hellinger merge threshold from 0.65 to 0.85 (ablation) | FROSS · threshold sweep | object recall: 69.8 -> 61.9 | Hurt |
| raise Hellinger merge threshold from 0.85 to 0.9 (ablation) | FROSS · threshold sweep | relationship recall: 30.0 -> 29.3 | Hurt |
| replace EGTR object detector with RT-DETR (ablation) | EGTR · ReplicaSSG | latency (ms): 14.6 -> 6.82 | Improved |
| replace EGTR object detector with RT-DETR (ablation) | EGTR · Visual Genome | R@50: 30.2 -> 22.0 | Hurt |
| replace EGTR object detector with RT-DETR (ablation) | EGTR · ReplicaSSG | AP@50: 21.2 -> 23.8 | Improved |
Later papers that compared with this method
2 later accepted papers report FROSS as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- DepWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors (ECCV 2026 · better on 8 of 8 shared results)
- PUF: Plug-and-Play Uncertainty-Aware Fusion for Online 3D Scene Graph Generation (ECCV 2026 · better on 7 of 8 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
- DepWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors (ECCV 2026 · similarity 0.91)
- Open3DSG: Open-Vocabulary 3D Scene Graphs from Point Clouds with Queryable Objects and Open-Set Relationships (CVPR 2024 · similarity 0.88)
- SGFormer: Semantic Graph Transformer for Point Cloud-Based 3D Scene Graph Generation (AAAI 2024 · similarity 0.87)
- Graph-Guided Scene Reconstruction from Images with 3D Gaussian Splatting (ICLR 2025 · similarity 0.87)
- Bootstraping Clustering of Gaussians for View-consistent 3D Scene Understanding (AAAI 2025 · similarity 0.87)
- Tackling View-Dependent Semantics in 3D Language Gaussian Splatting (ICML 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.
