Research map / Graph learning / Knowledge and temporal graphs
Constrained Structure Learning for Scene Graph Generation
Daqi Liu, Miroslaw Z. Bober, Josef Kittler
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 |
|---|---|---|---|
| use CSL instead of BGNN | CSL · Visual Genome SGDet | mR@50/100: 10.7 -> 11.9, 12.6 -> 14.3 | Improved |
| use CSL instead of BGNN | CSL · Visual Genome SGCls | mR@50/100: 14.3 -> 16.7, 16.5 -> 17.9 | Improved |
| use CSL instead of BGNN | CSL · Visual Genome PredCls | mR@50/100: 30.4 -> 29.5, 32.9 -> 31.6 | Hurt |
| use CSL instead of BGNN on tail predicates | CSL · Visual Genome SGDet R@100 Tail | R@100: 6.4 -> 8.8 | Improved |
| add balance adjustment strategy to CSL | CSL · Visual Genome PredCls | mR@50/100: 29.5 -> 36.9, 31.6 -> 39.2 | Improved |
| add balance adjustment strategy to CSL | CSL · Visual Genome SGDet | mR@50/100: 11.9 -> 15.7, 14.3 -> 18.4 | Improved |
| use CSL+BA instead of Transformer+BA | CSL+BA · Visual Genome PredCls | mR@50: 31.9 -> 36.9 | Improved |
| increase EMD iterations T from 5 to 10 (ablation) | CSL · Visual Genome SGDet | mR@100: 13.92 -> 14.44 | Improved |
| increase EMD iterations T from 10 to 20 (ablation) | CSL · Visual Genome SGDet | mR@100: 14.44 -> 14.34 | No clear effect |
| use CSL instead of BGNN | CSL · Open Images V6 | mR@50: 40.45 -> 41.72 | Improved |
| use CSL instead of BGNN | CSL · Open Images V6 | score_wtd: 42.06 -> 42.86 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GPS-Net* — GPS-Net: Graph Property Sensing Network for Scene Graph Generation (CVPR 2020) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Importance Weighted Structure Learning for Scene Graph Generation (TPAMI 2023 · similarity 0.93)
- Energy-Based Learning for Scene Graph Generation (CVPR 2021 · similarity 0.88)
- Conformal Prediction and MLLM aided Uncertainty Quantification in Scene Graph Generation (CVPR 2025 · similarity 0.87)
- Scene Graph Disentanglement and Composition for Generalizable Complex Image Generation (NeurIPS 2024 · similarity 0.87)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.86)
- Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency (ICML 2024 · similarity 0.85)
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.
