Research map / Graph learning / Knowledge and temporal graphs
Unconditional Scene Graph Generation
Sarthak Garg, Helisa Dhamo, Azade Farshad, Sabrina Musatian, Nassir Navab, Federico Tombari
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 node information and E_from/E_to conditioning to edge generation (SceneGraphGen vs GraphRNN, BFS) (ablation) | SceneGraphGen · Visual Genome graphs | MMD graph (x10^3): 1.3 -> 1.82 | Mixed |
| use SceneGraphGen instead of GraphRNN with random ordering | GraphRNN · Visual Genome graphs | MMD graph (x10^3): 1.2 -> 0.11 | Improved |
| use SceneGraphGen instead of GraphRNN with random ordering | GraphRNN · Visual Genome graphs | MMD node (x10^3): 0.39 -> 0.37 | No clear effect |
| use random node ordering instead of BFS ordering (ablation) | SceneGraphGen · Visual Genome graphs | MMD graph (x10^3): 1.82 -> 0.11 | Improved |
| use random node ordering instead of hierarchical ordering (ablation) | SceneGraphGen · Visual Genome graphs | MMD graph (x10^3): 0.63 -> 0.11 | Improved |
| use random node ordering instead of BFS ordering (ablation) | SceneGraphGen · Visual Genome graphs | MMD node (x10^3): 2.05 -> 0.37 | Improved |
| use random node ordering instead of hierarchical ordering (ablation) | SceneGraphGen · Visual Genome images | FID: 72.2 -> 71.2 | Improved |
| use random node ordering instead of hierarchical ordering (ablation) | SceneGraphGen · Visual Genome images | IS: 5.26 -> 4.95 | Hurt |
| use SceneGraphGen graphs instead of GraphRNN graphs to drive sg2im | GraphRNN · Visual Genome images | FID: 74.5 -> 71.2 | Improved |
| use scene graphs plus sg2im instead of unconditional StyleGAN2 | StyleGAN2 · Visual Genome images | FID: 66.3 -> 71.2 | Hurt |
| use scene graphs plus sg2im instead of unconditional StyleGAN2 | StyleGAN2 · Visual Genome images | detected object categories: 40 -> 50 | Improved |
| use scene graphs plus sg2im instead of unconditional StyleGAN2 | StyleGAN2 · Visual Genome images | object occurrence error vs ground truth: 1.4 -> 1.2 | Improved |
Closest papers by meaning
- Scene Graph Generation by Iterative Message Passing (CVPR 2017 · similarity 0.94)
- Image Generation From Scene Graphs (CVPR 2018 · similarity 0.93)
- IS-GGT: Iterative Scene Graph Generation With Generative Transformers (CVPR 2023 · similarity 0.92)
- Learning Canonical Representations for Scene Graph to Image Generation (ECCV 2020 · similarity 0.92)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.90)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.90)
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.
