Research map / Graph learning / Knowledge and temporal graphs
SGTR: End-to-End Scene Graph Generation With Transformer
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SGTR rather than ResNet-101 BGNN | OpenImage V6 · OpenImage V6 | wmAP rel: 31.15 -> 36.98 | Helped |
| use SGTR rather than ResNet-101 BGNN | OpenImage V6 · OpenImage V6 | wmAP phr: 31.37 -> 38.73 | Helped |
| add entity-aware predicate node representations (ablation) | SGTR · Visual Genome validation | mR@100: 15.9 -> 17.3 | Helped |
| use the structural predicate decoder instead of shared cross-attention (ablation) | SGTR · Visual Genome validation | mR@100: 15.1 -> 17.3 | Helped |
| add entity indicators to predicate decoding (ablation) | SGTR · Visual Genome validation | mR@100: 14.8 -> 17.3 | Helped |
| use graph assembling rather than indicator predictions as entity nodes (ablation) | SGTR · Visual Genome validation | mR@100: 7.0 -> 17.3 | Helped |
| increase predicate and entity-indicator decoder depth from 3 to 6 layers (ablation) | SGTR · Visual Genome validation | mR@100: 13.3 -> 17.3 | Helped |
| increase predicate and entity-indicator decoder depth from 6 to 12 layers (ablation) | SGTR · Visual Genome validation | mR@100: 17.3 -> 17.0 | Hurt |
| use joint semantic and spatial similarity for graph assembling rather than spatial-only matching (ablation) | SGTR · Visual Genome validation | mR@100: 11.8 -> 17.3 | Helped |
| use joint semantic and spatial similarity for graph assembling rather than feature-only matching (ablation) | SGTR · Visual Genome validation | mR@100: 16.1 -> 17.3 | Helped |
| apply bi-level resampling (ablation) | SGTR · Visual Genome test, SGDet | mR@100: 15.2 -> 20.1 | Helped |
Closest papers by meaning
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.92)
- Leveraging Predicate and Triplet Learning for Scene Graph Generation (CVPR 2024 · similarity 0.90)
- From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models (CVPR 2024 · similarity 0.89)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.89)
- SGFormer: Semantic Graph Transformer for Point Cloud-Based 3D Scene Graph Generation (AAAI 2024 · similarity 0.88)
- EGTR: Extracting Graph from Transformer for Scene Graph Generation (CVPR 2024 · 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.