Research map / Graph learning / Knowledge and temporal graphs
Environment-Invariant Curriculum Relation Learning for Fine-Grained Scene Graph Generation
Yukuan Min, 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 EICR to VCTree | VCTree · VG, three tasks | mR@50/100 and F@50/100: PredCls F@50/100 26.6 / 28.6 -> 43.6/ 45.8 | Improved |
| add EICR to Motifs | Motifs · VG PredCls | R@50/100: 65.2 / 67.0 -> 55.3 / 57.4 | Hurt |
| add curriculum schedule to three environments (ablation) | EICR on Motifs · VG SGCls | mR@50/100: 15.4 / 16.1 -> 20.8 / 21.8 | Improved |
| add curriculum schedule to three environments (ablation) | EICR on Motifs · VG SGCls | R@50/100: 39.3 / 40.1 -> 34.5 / 35.4 | Hurt |
| combine normal, balanced and over-balanced environments instead of two (ablation) | EIL on Motifs · VG SGCls | F@50/100: 21.8 / 23.0 -> 22.1 / 23.0 | Mixed |
| add environment-invariant learning alone to Motifs (ablation) | Motifs · VG SGCls | mR@50/100: 8.3 / 8.8 -> 15.4 / 16.1 | Improved |
| add IRM regularization term (ablation) | EICR on Motifs · VG PredCls | F@50/100: 42.2 / 44.3 -> 42.8 / 45.0 | Improved |
| add IRM regularization term (ablation) | EICR on Motifs · VG PredCls | mR@50/100: 35.4 / 37.4 -> 34.9 / 37.0 | Hurt |
| increase lambda_max from 0.9 to 0.99 (ablation) | EICR on Motifs · VG SGCls | F@50/100: 25.9 / 27.0 -> 23.8 / 24.8 | Hurt |
| remove curriculum schedule for over-balanced environment (ablation) | EICR on Motifs · VG SGCls | F@50/100: 24.0 / 25.1 -> 25.9 / 27.0 | Improved |
Closest papers by meaning
- From General to Specific: Informative Scene Graph Generation via Balance Adjustment (ICCV 2021 · similarity 0.89)
- Semantic Diversity-aware Prototype-based Learning for Unbiased Scene Graph Generation (ECCV 2024 · similarity 0.89)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.88)
- Leveraging Predicate and Triplet Learning for Scene Graph Generation (CVPR 2024 · similarity 0.87)
- Hydra-SGG: Hybrid Relation Assignment for One-stage Scene Graph Generation (ICLR 2025 · similarity 0.87)
- RA-SGG: Retrieval-Augmented Scene Graph Generation Framework via Multi-Prototype Learning (AAAI 2025 · 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.
