Research map / Graph learning / Knowledge and temporal graphs
Robo-SGG: Exploiting Layout-Oriented Normalization and Restitution Can Improve Robust Scene Graph Generation
Changsheng Lv, Zijian Fu, Mengshi Qi
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 Robo-SGG (NRM + LEE) to HiKER | HiKER · VG-C SGCls | mR@50: 3.5 -> 3.7 | Improved |
| add Robo-SGG to MOTIFS | MOTIFS · VG clean SGDet | mR@50: 7.2 -> 7.1 | No clear effect |
| add Robo-SGG to VTransE | VTransE · VG clean SGCls | mR@50: 12.6 -> 12.4 | Hurt |
| use NRM instead of SNR restitution (with LEE) (ablation) | MOTIFS · VG-C PredCls | mR@100: +2.2% -> +4.3% | Improved |
| add LEE alone (ablation) | HiKER · VG-C SGCls | mR@50: 3.5 -> 3.5 | No clear effect |
| add NRM alone (ablation) | MOTIFS · VG-C PredCls | mR@50: 12.6 -> 12.8 | Improved |
| add both LEE and NRM (ablation) | MOTIFS · VG-C PredCls | mR@50: 12.6 -> 13.8 | Improved |
| fuse bbox embedding by concatenation instead of gating (ablation) | VCTree LEE · VG-C SGDet | mR@50: 2.51 -> 2.46 | Hurt |
| fuse bbox embedding by addition instead of gating (ablation) | VCTree LEE · VG-C SGDet | mR@50: 2.51 -> 2.48 | Hurt |
| add gate fusion in LEE (ablation) | VCTree · VG-C SGDet | mR@50: 2.51 -> 2.62 | Improved |
| use object centroids rather than bounding boxes in layout-aware attention (ablation) | VCTree NRM · VG-C SGDet | mR@50: 2.55 -> 2.69 | Improved |
| add Robo-SGG under increasing Gaussian noise severity | baseline SGG models · VG-C SGCls Gaussian noise | mR@100: 4.6% (severity 1) -> 28.2% (severity 5) relative gain | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- RelTR — RelTR: Relation Transformer for Scene Graph Generation (TPAMI 2023) · better on 1 of 1 shared result
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
- HiKER-SGG: Hierarchical Knowledge Enhanced Robust Scene Graph Generation (CVPR 2024 · similarity 0.90)
- Towards Unbiased and Robust Spatio-Temporal Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.87)
- Conformal Prediction and MLLM aided Uncertainty Quantification in Scene Graph Generation (CVPR 2025 · similarity 0.86)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.86)
- Adaptive Visual Scene Understanding: Incremental Scene Graph Generation (NeurIPS 2024 · similarity 0.85)
- TD²-Net: Toward Denoising and Debiasing for Video Scene Graph Generation (AAAI 2024 · similarity 0.85)
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.
