Research map / Graph learning / Knowledge and temporal graphs
Vision Relation Transformer for Unbiased Scene Graph Generation
Gopika Sudhakaran, Devendra Singh Dhami, Kristian Kersting, Stefan Roth
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 MEET mutually exclusive experts on top of VETO (ablation) | VETO · VG150 PredCls | A@100: 45.5 -> 65.7 | Improved |
| apply MEET to Motifs | Motifs · VG150 PredCls | A@100: 41.6 -> 53.1 | Improved |
| apply MEET to SHA | SHA · VG150 PredCls | A@100: 43.5 -> 54.3 | Improved |
| apply MEET to VTransE | VTransE · VG150 SGDet | A@100: 20.2 -> 17.6 | Hurt |
| apply MEET to Motifs | Motifs · VG150 SGDet | A@100: 22.6 -> 22.6 | No clear effect |
| add local-level entity patch generation (ablation) | VETO · VG150 SGCls | A@100: 20.9 -> 25.7 | Improved |
| add depth maps without cross-modality fusion (ablation) | VETO · VG150 SGCls | mR@100: 15.1 -> 14.1 | Hurt |
| add cross-modality patch fusion with depth (ablation) | VETO · VG150 SGCls | A@100: 25.3 -> 25.9 | Improved |
| add cross-relation patch generation (ablation) | VETO · VG150 SGCls | A@100: 25.7 -> 26.4 | Improved |
| use higher-quality depth maps (VG-Depth.v2) (ablation) | VETO · VG PredCls | mR@100: 33.5 -> 35.1 | Improved |
| use higher-quality depth maps (VG-Depth.v2) (ablation) | Depth-VRD · VG PredCls | mR@100: 23.1 -> 23.1 | No clear effect |
| use weaker object detector (OD1) in SGDet (ablation) | VETO vs Motifs · VG SGDet | A@50 drop (%): Motifs 3.3 vs VETO 5.7 | Hurt |
Closest papers by meaning
- RelTR: Relation Transformer for Scene Graph Generation (TPAMI 2023 · similarity 0.91)
- EGTR: Extracting Graph from Transformer for Scene Graph Generation (CVPR 2024 · similarity 0.91)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.90)
- Hydra-SGG: Hybrid Relation Assignment for One-stage Scene Graph Generation (ICLR 2025 · similarity 0.90)
- Leveraging Predicate and Triplet Learning for Scene Graph Generation (CVPR 2024 · similarity 0.89)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.88)
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.
