Research map / Graph learning / Knowledge and temporal graphs
Graph R-CNN for Scene Graph Generation
Jianwei Yang, Jiasen Lu, Stefan Lee, Dhruv Batra, Devi Parikh
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 the relation proposal network (RePN) (ablation) | Graph R-CNN baseline · Visual Genome | SGGen R@50: 6.1 -> 8.7 | Improved |
| add RePN (ablation) | Graph R-CNN baseline · Visual Genome | SGGen+ R@100: 27.9 -> 34.8 | Improved |
| add RePN (ablation) | Graph R-CNN baseline · Visual Genome | detection mAP@0.5: 20.4 -> 23.6 | Improved |
| add RePN (ablation) | Graph R-CNN baseline · Visual Genome (ground-truth boxes) | PredCls R@50: 33.5 -> 34.5 | No clear effect |
| add plain GCN on top of RePN (ablation) | Graph R-CNN with RePN · Visual Genome | PredCls R@50: 34.5 -> 52.3 | Improved |
| add plain GCN on top of RePN (ablation) | Graph R-CNN with RePN · Visual Genome | detection mAP@0.5: 23.6 -> 23.4 | No clear effect |
| use attentional GCN instead of plain GCN (ablation) | Graph R-CNN with RePN · Visual Genome | SGGen R@50: 10.8 -> 11.4 | Improved |
| use attentional GCN instead of plain GCN (ablation) | Graph R-CNN with RePN · Visual Genome | PredCls R@50: 52.3 -> 54.2 | Improved |
| use attentional GCN instead of plain GCN (ablation) | Graph R-CNN with RePN · Visual Genome | detection mAP@0.5: 23.4 -> 23.0 | Hurt |
| use Graph R-CNN instead of reimplemented IMP | scene graph generation · Visual Genome | SGGen R@50: 6.4 -> 11.4 | Improved |
| use Graph R-CNN instead of frequency baseline NM-Freq | scene graph generation · Visual Genome | SGGen+ R@100: 27.8 -> 35.9 | Improved |
| use SGGen+ instead of SGGen when 20% of objects with relationships are perturbed | evaluation metric · perturbed ground-truth graphs | recall: 54.1 vs 84.3 | Mixed |
Closest papers by meaning
- LinkNet: Relational Embedding for Scene Graph (NeurIPS 2018 · similarity 0.91)
- Generating Triples With Adversarial Networks for Scene Graph Construction (AAAI 2018 · similarity 0.91)
- Structured Sparse R-CNN for Direct Scene Graph Generation (CVPR 2022 · similarity 0.90)
- Scene Graph Generation by Iterative Message Passing (CVPR 2017 · similarity 0.90)
- Image Generation From Scene Graphs (CVPR 2018 · similarity 0.90)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.89)
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.
