Research map / Graph learning / Knowledge and temporal graphs
Learning Canonical Representations for Scene Graph to Image Generation
Roei Herzig, Amir Bar, Huijuan Xu, Gal Chechik, Trevor Darrell, Amir Globerson
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 weighted scene graph canonicalization (WSGC) | Sg2Im 5 GCN layout model · Packed COCO | mIOU: 35.8 -> 39.3 | Improved |
| add weighted scene graph canonicalization (WSGC) | Sg2Im 5 GCN layout model · Packed VG | mIOU: 25.4 -> 28.5 | Improved |
| add WSGC | Sg2Im 5 GCN · Packed COCO | R@0.5: 25.3 -> 30.1 | Improved |
| add WSGC | Sg2Im 5 GCN · Standard COCO | mIOU: 41.7 -> 41.9 | No clear effect |
| add WSGC | Sg2Im 5 GCN · Standard VG | mIOU: 16.9 -> 18.0 | No clear effect |
| increase Sg2Im GCN depth from 5 to 16 layers (ablation) | Sg2Im · Packed COCO | mIOU: 35.8 -> 37.7 | Mixed |
| increase Sg2Im GCN depth from 8 to 16 layers (ablation) | Sg2Im · Packed COCO semantically equivalent SGs | mIOU: 23.6 -> 21.6 | Hurt |
| add WSGC | Sg2Im 5 GCN · Packed COCO semantically equivalent SGs | mIOU: 21.8 -> 35.3 | Improved |
| add WSGC | Sg2Im 5 GCN · Packed COCO noisy SGs (10% relations modified) | mIOU: 29.4 -> 31.8 | Improved |
| use exact WSGC-E instead of sampling WSGC-S (ablation) | WSGC 5 GCN · COCO standard | mIOU: 41.9 -> 42.2 | No clear effect |
| use exact WSGC-E instead of sampling WSGC-S (ablation) | WSGC 5 GCN · Visual Genome standard | R@0.5: 10.6 -> 9.9 | No clear effect |
| add WSGC layouts to LostGAN image generation | LostGAN · COCO | Inception score: 5.4 -> 5.6 | Improved |
Closest papers by meaning
- Scene Graph Generation by Iterative Message Passing (CVPR 2017 · similarity 0.93)
- Image Generation From Scene Graphs (CVPR 2018 · similarity 0.92)
- Unconditional Scene Graph Generation (ICCV 2021 · similarity 0.92)
- Scene Graph Disentanglement and Composition for Generalizable Complex Image Generation (NeurIPS 2024 · similarity 0.90)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.89)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.89)
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.
