Research map / Graph learning / Knowledge and temporal graphs
Deep Generative Probabilistic Graph Neural Networks for Scene Graph Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the sequential DG-PGNN node addition and message passing over the initial complete PGN (ablation) | initial complete PGN · Visual Genome PRED-CLS | R@50: 63.5 -> 70.1 | Helped |
| add the sequential DG-PGNN node addition and message passing over the initial complete PGN (ablation) | initial complete PGN · Visual Genome SG-GEN | R@50: 26.1 -> 32.1 | Helped |
| add region-grounded caption-guided updates (ablation) | DG-PGNN · Visual Genome PRED-CLS | R@50: 67.3 -> 70.1 | Helped |
| add region-grounded caption-guided updates (ablation) | DG-PGNN · Visual Genome SG-CLS | R@50: 38.6 -> 39.5 | Helped |
| add region-grounded caption-guided updates (ablation) | DG-PGNN · Visual Genome SG-GEN | R@50: 30.2 -> 32.1 | Helped |
| replace VGG backbone features with ResNet (ablation) | DG-PGNN · Visual Genome PRED-CLS | R@50: 69.0 -> 70.1 | No clear effect |
| replace VGG backbone features with ResNet (ablation) | DG-PGNN · Visual Genome SG-GEN | R@50: 31.2 -> 32.1 | No clear effect |
| add captions to scene graph used for VQA (ablation) | DG-PGNN · Visual7W telling | avg accuracy: 68.0 -> 69.3 | Helped |
| use predicted scene graph attention instead of MLP scoring | VQA model · Visual7W telling | avg accuracy: 67.1 -> 69.3 | Helped |
| use DG-PGNN scene graphs rather than program supervision | VQA model · CLEVR test | overall accuracy: 96.9 -> 96.1 | No clear effect |
Closest papers by meaning
- Generating Triples With Adversarial Networks for Scene Graph Construction (AAAI 2018 · similarity 0.92)
- IS-GGT: Iterative Scene Graph Generation With Generative Transformers (CVPR 2023 · similarity 0.91)
- From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models (CVPR 2024 · similarity 0.89)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.88)
- Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency (ICML 2024 · similarity 0.88)
- Scene Graph-Grounded Image Generation (AAAI 2025 · 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.