Research map / Graph learning / Knowledge and temporal graphs
End-to-End Entity-Predicate Association Reasoning for Dynamic Scene Graph Generation
Liwei Wang, Yanduo Zhang, Tao Lu, Fang Liu, Huiqin Zhang, Jiayi Ma, Huabing Zhou
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 focal loss instead of cross-entropy for predicate prediction (ablation) | DETR-based DSGG baseline · Action Genome SGDet (No Constraint) | R@10: 28.4 -> 31.7 | Improved |
| use focal loss instead of cross-entropy for predicate prediction (ablation) | DETR-based DSGG baseline · Action Genome SGDet (No Constraint) | mR@10: 19.6 -> 25.6 | Improved |
| add CLIP text-embedding triplet cue modeling (ablation) | ARN with focal loss · Action Genome SGDet (No Constraint) | R@50: 45.4 -> 51.7 | Improved |
| add CLIP text-embedding triplet cue modeling (ablation) | ARN with focal loss · Action Genome SGDet (No Constraint) | mR@10: 25.6 -> 27.1 | Improved |
| add Predicate Association Parsing module (ablation) | ARN without PAP and HA · Action Genome SGDet (No Constraint) | R@10: 34.4 -> 35.3 | Improved |
| add Predicate Association Parsing module (ablation) | ARN without PAP and HA · Action Genome SGDet (No Constraint) | mR@50: 48.5 -> 49.8 | Improved |
| add Hierarchical Attention on top of PAP (ablation) | ARN · Action Genome SGDet (No Constraint) | R@10: 35.3 -> 37.6 | Improved |
| add Hierarchical Attention on top of PAP (ablation) | ARN · Action Genome SGDet (No Constraint) | mR@50: 49.8 -> 51.9 | Improved |
| add paired entity features directly to predicate query instead of conditional weight mapping (Model I) (ablation) | ARN · Action Genome SGDet (No Constraint) | R@10: 37.6 -> 35.0 | Hurt |
| use nonlinear mapping instead of conditional weight mapping (Model II) (ablation) | ARN · Action Genome SGDet (No Constraint) | mR@10: 29.7 -> 29.3 | Hurt |
| add conditional weight mapping instead of nonlinear mapping (ablation) | ARN · Action Genome SGDet (No Constraint) | R@50: 53.0 -> 54.1 | Improved |
| use ARN versus OED under constraint at R@50 | ARN vs OED · Action Genome SGDet (With Constraint) | R@50: 48.9 -> 47.2 | Hurt |
Closest papers by meaning
- Leveraging Predicate and Triplet Learning for Scene Graph Generation (CVPR 2024 · similarity 0.90)
- A Closer Look at Dynamic Scene Graph Generation in the Era of Multimodal Large Language Models (NeurIPS 2026 · similarity 0.90)
- OED: Towards One-stage End-to-End Dynamic Scene Graph Generation (CVPR 2024 · similarity 0.89)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.88)
- HyperGLM: HyperGraph for Video Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.88)
- Synthetic Visual Genome (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.
