Research map / Graph learning / Knowledge and temporal graphs
Mixture-of-Experts based Feature Decoupling for Open Vocabulary Scene Graph Generation
Yiming Li, Sisi You, Bing-Kun Bao
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 object feature decoupling via MoE (ablation) | MoE-FD · OvD+R-SGG VG150, Joint Base+Novel | R@20: 15.72 -> 17.35 | Improved |
| add relationship feature decoupling via MoE (ablation) | MoE-FD · OvD+R-SGG VG150, Joint Base+Novel | R@20: 17.03 -> 17.35 | Improved |
| add iterative feature refinement between object and relation features (ablation) | MoE-FD · OvD+R-SGG VG150, Joint Base+Novel | R@20: 16.09 -> 17.35 | Improved |
| use 8 object experts and 6 relation experts (ablation) | MoE-FD · OvD+R-SGG VG150 R@20 | R@20 | Improved |
| set knowledge selection threshold epsilon to 0.7 (ablation) | MoE-FD · OvD+R-SGG VG150 R@20 | R@20 | Improved |
| use MoE-FD instead of OvSGTR (Swin-B) | MoE-FD · OvD-SGG VG150, Novel Obj. | R@50: 15.58 -> 20.94 | Improved |
| use MoE-FD instead of ACC (Swin-B) | MoE-FD · OvD+R-SGG VG150, Novel Rel. | R@20: 12.37 -> 16.61 | Improved |
| use larger Swin-B backbone instead of Swin-T | MoE-FD · OvR-SGG VG150, Novel Rel. | R@20: 16.01 -> 18.06 | Improved |
Closest papers by meaning
- Interaction-Centric Knowledge Infusion and Transfer for Open Vocabulary Scene Graph Generation (NeurIPS 2025 · similarity 0.88)
- Scene Graph Generation with Role-Playing Large Language Models (NeurIPS 2024 · similarity 0.88)
- Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation (AAAI 2025 · similarity 0.87)
- From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models (CVPR 2024 · similarity 0.86)
- Synthetic Visual Genome (CVPR 2025 · similarity 0.86)
- Towards Flexible Visual Relationship Segmentation (NeurIPS 2024 · similarity 0.86)
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.
