Research map / Probabilistic and scientific modeling / Variational and generative sampling
Importance Weighted 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 |
|---|---|---|---|
| replace message passing ELBO with importance weighted bound and entropic mirror descent (IWSL vs BGNN) | SGG on Visual Genome · VG SGDet | mR@50: 10.7 -> 13.7 | Improved |
| IWSL vs BGNN | SGG on Visual Genome · VG SGCls | mR@50: 14.3 -> 17.4 | Improved |
| IWSL vs BGNN | SGG on Visual Genome · VG PredCls | mR@50: 30.4 -> 30.0 | No clear effect |
| IWSL vs BGNN on tail predicates | SGG on Visual Genome · VG SGDet R@100 Tail | recall: 6.4 -> 10.7 | Improved |
| IWSL vs BGNN on head predicates | SGG on Visual Genome · VG SGDet R@100 Head | recall: 33.4 -> 30.6 | Hurt |
| add balance adjustment strategy to IWSL | IWSL · VG PredCls | mR@50: 30.0 -> 36.9 | Improved |
| IWSL+BA vs Transformer+BA | SGG with BA · VG PredCls | mR@100: 34.2 -> 39.2 | Improved |
| IWSL vs BGNN on Open Images | SGG on Open Images V6 · Open Images V6 | mR@50: 40.45 -> 42.18 | Improved |
| IWSL vs BGNN on weighted score | SGG on Open Images V6 · Open Images V6 | score_wtd: 42.06 -> 41.87 | No clear effect |
| increase number of samples in variational inference from 10 to 30 (ablation) | IWSL · VG SGDet | mR@100: 15.5 -> 15.6 | No clear effect |
| increase number of samples from 10 to 50 (ablation) | IWSL · VG SGDet | mR@50: 13.1 -> 13.7 | Improved |
| increase number of samples from 10 to 50 (ablation) | IWSL · VG SGDet | mR@20: 9.9 -> 10.6 | Improved |
Closest papers by meaning
- Constrained Structure Learning for Scene Graph Generation (TPAMI 2023 · similarity 0.93)
- Conformal Prediction and MLLM aided Uncertainty Quantification in Scene Graph Generation (CVPR 2025 · similarity 0.88)
- Energy-Based Learning for Scene Graph Generation (CVPR 2021 · similarity 0.87)
- Neural Belief Propagation for Scene Graph Generation (TPAMI 2023 · similarity 0.86)
- Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency (ICML 2024 · similarity 0.86)
- Scene Graph Disentanglement and Composition for Generalizable Complex Image Generation (NeurIPS 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.
