Research map / Vision-language and multimodal / Human and embodied modeling
Vision-Language Interactive Relation Mining for Open-Vocabulary Scene Graph Generation
Yukuan Min, Muli Yang, Jinhao Zhang, Yuxuan Wang, Aming Wu, Cheng Deng
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 learnable context tokens (ablation) | VL-IRM · VG150 Novel+base | R@100: 16.67 -> 18.48 | Improved |
| add learnable context tokens (vs frozen) (ablation) | VL-IRM · VG150 Novel+base | R@100: 15.36 -> 18.48 | Improved |
| add dynamic fitting debiasing (ablation) | VL-IRM · VG150 Novel+base | R@100: 16.24 -> 18.48 | Improved |
| add dynamic fitting debiasing (ablation) | VL-IRM · VG150 Novel | R@100: 10.23 -> 13.45 | Improved |
| use dynamic fitting instead of static LA (ablation) | VL-IRM · VG150 Novel+base | R@100: 12.58 -> 18.48 | Improved |
| use dynamic fitting instead of PGLA (ablation) | VL-IRM · VG150 Novel+base | R@100: 13.42 -> 18.48 | Improved |
| add background and foreground pseudo labels vs foreground only (ablation) | VL-IRM · VG150 Novel+Base | R@100: 19.12 -> 19.71 | Improved |
| add pseudo labeling (Bg & fg) (ablation) | VL-IRM · VG150 Novel | R@100: 13.45 -> 14.32 | Improved |
| add relation language generator (lambda>0) (ablation) | VL-IRM · VG150 Novel | performance | Improved |
| combine VL-IRM with OVSGTR | OVSGTR · VG SGDet extra data, Novel+base | R@100: 23.9 -> 25.0 | Improved |
| combine VL-IRM with OVSGTR-T | OVSGTR-T · VG150 OvD+R-SGG Novel relation | R@100: 11.19 -> 15.26 | Improved |
| VL-IRM vs OVSGTR on OvD+R-SGG novel R@50 | OVSGTR-T · VG150 OvD+R-SGG Novel | R@50: 9.20 -> 12.27 | Improved |
Closest papers by meaning
- Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation (AAAI 2025 · similarity 0.92)
- From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models (CVPR 2024 · similarity 0.91)
- Learning To Generate Language-Supervised and Open-Vocabulary Scene Graph Using Pre-Trained Visual-Semantic Space (CVPR 2023 · similarity 0.91)
- Scene Graph Generation with Role-Playing Large Language Models (NeurIPS 2024 · similarity 0.90)
- Interaction-Centric Knowledge Infusion and Transfer for Open Vocabulary Scene Graph Generation (NeurIPS 2025 · similarity 0.89)
- Leveraging Predicate and Triplet Learning 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.
