Research map / Label-efficient and robust learning / Recognition and long-tail
Exploring Long Tail Visual Relationship Recognition With Large Vocabulary
Sherif Abdelkarim, Aniket Agarwal, Panos Achlioptas, Jun Chen, Jiaji Huang, Boyang Li, Kenneth Church, Mohamed Elhoseiny
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 hub100k to the baseline classifier | baseline · GQA triplet | overall triplet accuracy: 14.7 -> 15.3 | Improved |
| add hub100k to the baseline classifier | baseline · VG8K triplet | overall triplet accuracy: 14.2 -> 15.4 | Improved |
| add hub100k to the baseline classifier | baseline · GQA triplet, thresholds (20, 100) with synsets | few-shot triplet score: 12.6 -> 15.5 | Improved |
| add hub100k to the baseline classifier | baseline · GQA triplet, thresholds (20, 100) with synsets | many-shot triplet score: 63.2 -> 59.8 | Hurt |
| increase hub size from hub1k to hub100k (ablation) | hub variants · GQA triplet, thresholds (20, 100) with synsets | all-class triplet score: 13.1 -> 16.0 | Improved |
| increase hub size from hub100k to hub500k (ablation) | hub variants · GQA triplet, thresholds (20, 100) with synsets | all-class triplet score: 16.0 -> 16.0 | No clear effect |
| add hub20k to weighted cross-entropy | w_ce · GQA triplet | overall triplet accuracy: 5.3 -> 4.7 | Hurt |
| add hub100k to dcpl | dcpl · GQA triplet | overall triplet accuracy: 11.1 -> 9.5 | Hurt |
| add hub100k to fc | fc · GQA triplet | overall triplet accuracy: 13.4 -> 14.0 | Improved |
| add hub20k to focal loss (gamma 5) | focal_5 · GQA triplet | overall triplet accuracy: 14.7 -> 15.3 | Improved |
| use weighted cross-entropy instead of baseline | baseline · GQA triplet | overall triplet accuracy: 14.7 -> 5.3 | Hurt |
| add hub10k to the baseline classifier | baseline · VG8K triplet | overall triplet accuracy: 14.2 -> 14.5 | Improved |
Closest papers by meaning
- RelTransformer: A Transformer-Based Long-Tail Visual Relationship Recognition (CVPR 2022 · similarity 0.86)
- Learning of Visual Relations: The Devil Is in the Tails (ICCV 2021 · similarity 0.85)
- Large-Scale Visual Relationship Understanding (AAAI 2019 · similarity 0.84)
- Visual Relationship Detection with Language Priors (ECCV 2016 · similarity 0.84)
- Towards Flexible Visual Relationship Segmentation (NeurIPS 2024 · similarity 0.83)
- Multi-Modal Prompting for Open-Vocabulary Video Visual Relationship Detection (AAAI 2024 · similarity 0.82)
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.
