Research map / Image generation / Autoregressive and latent generation
Exploiting Relationship for Complex-scene Image 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 |
|---|---|---|---|
| Use relationship-aware generation rather than sg2im | Complex-scene image generation · Visual Genome 64×64 | IS: 5.5 ± 0.1 -> 7.0 ± 0.2 | Helped |
| Use relationship-aware generation rather than sg2im | Complex-scene image generation · Visual Genome 128×128 | IS: 6.2 ± 0.2 -> 9.2 ± 0.8 | Helped |
| Use relationship-aware generation rather than sg2im | Complex-scene image generation · HICO-DET 64×64 | FID: 94.3 -> 47.4 | Helped |
| Use relationship-aware generation rather than sg2im | Complex-scene image generation · HICO-DET 128×128 | FID: 123.0 -> 61.4 | Helped |
| Add Pair-wise Spatial Constraint Module with relative scale and distance supervision (ablation) | Ours · Visual Genome 128×128 | IS: 8.6 ± 1.2 -> 9.2 ± 0.8 | Helped |
| Add Pair-wise Spatial Constraint Module with relative scale and distance supervision (ablation) | Ours · Visual Genome 128×128 | FID: 59.8 -> 53.0 | Helped |
| Add relation embeddings to object shape and texture features (ablation) | Ours · Visual Genome 128×128 | IS: 8.7 ± 0.9 -> 9.2 ± 0.8 | Helped |
| Add relation embeddings to object shape and texture features (ablation) | Ours · Visual Genome 128×128 | FID: 57.4 -> 53.0 | Helped |
| Add Scene Graph Discriminator loss (ablation) | Ours · Visual Genome 128×128 | IS: 7.4 ± 0.2 -> 9.2 ± 0.8 | Helped |
| Add Scene Graph Discriminator loss (ablation) | Ours · Visual Genome 128×128 | FID: 73.3 -> 53.0 | Helped |
Closest papers by meaning
- Object-Centric Image Generation from Layouts (AAAI 2021 · similarity 0.92)
- Image Generation From Scene Graphs (CVPR 2018 · similarity 0.92)
- Generating Multiple Objects at Spatially Distinct Locations (ICLR 2019 · similarity 0.91)
- Context-Aware Layout to Image Generation With Enhanced Object Appearance (CVPR 2021 · similarity 0.90)
- Scene Graph Disentanglement and Composition for Generalizable Complex Image Generation (NeurIPS 2024 · similarity 0.89)
- 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.