Research map / Architectures and efficiency / Attention and transformers
RelTransformer: A Transformer-Based Long-Tail Visual Relationship Recognition
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 RelTransformer rather than LSVRU with DCPL | VG8K-LT relation prediction · VG8K-LT | average per-class accuracy (all): 14.4 -> 19.2 | Helped |
| Use RelTransformer rather than LSVRU with WCE | GQA-LT relation prediction · GQA-LT | average per-class accuracy (all): 20.5 -> 46.5 | Helped |
| Use RelTransformer rather than RTN | VG200 predicate classification · VG200 | R@50: 68.7 -> 69.7 | Helped |
| Use RelTransformer rather than Motif + EBM | VG200 predicate classification · VG200 | mR@50: 18.2 -> 19.58 | Helped |
| Add global-context encoder (ablation) | RelTransformer with CE · VG8K-LT | average per-class accuracy (all): 14.8 -> 16.1 | Helped |
| Add memory attention (ablation) | RelTransformer with CE · VG8K-LT | average per-class accuracy (all): 15.0 -> 16.1 | Helped |
| Add global-context encoder (ablation) | RelTransformer with Focal Loss · VG8K-LT | average per-class accuracy (all): 15.1 -> 16.8 | Helped |
| Add memory attention (ablation) | RelTransformer with Focal Loss · VG8K-LT | average per-class accuracy (all): 15.9 -> 16.8 | Helped |
| Add global-context encoder (ablation) | RelTransformer with DCPL · VG8K-LT | average per-class accuracy (all): 18.2 -> 19.2 | Helped |
| Add memory attention (ablation) | RelTransformer with DCPL · VG8K-LT | average per-class accuracy (all): 18.8 -> 19.2 | Helped |
| Add global-context encoder (ablation) | RelTransformer with WCE · VG8K-LT | average per-class accuracy (all): 17.6 -> 19.0 | Helped |
| Add memory attention (ablation) | RelTransformer with WCE · VG8K-LT | average per-class accuracy (all): 18.7 -> 19.0 | Helped |
Closest papers by meaning
- VRDFormer: End-to-End Video Visual Relation Detection With Transformers (CVPR 2022 · similarity 0.89)
- Groupwise Query Specialization and Quality-Aware Multi-Assignment for Transformer-based Visual Relationship Detection (CVPR 2024 · similarity 0.88)
- EGTR: Extracting Graph from Transformer for Scene Graph Generation (CVPR 2024 · similarity 0.88)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.87)
- Towards Flexible Visual Relationship Segmentation (NeurIPS 2024 · similarity 0.87)
- Leveraging Predicate and Triplet Learning for Scene Graph Generation (CVPR 2024 · 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.