Research map / Image generation / Text-to-image diffusion
Advancing Aesthetic Image Generation via Composition Transfer
Kai Zou, Zhiwei Zhao, Bin Liu, Nenghai Yu
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 LVLM-based composition planning instead of semantic filtering only (ablation) | Composer · MS-COCO | ImageReward: 0.91 -> 1.09 | Improved |
| add LVLM-based composition planning instead of semantic filtering only (ablation) | Composer · MS-COCO | Aesthetic score: 5.82 -> 5.93 | Improved |
| use Union-ControlNet instead of Dual-ControlNet (ablation) | Composer · SD-v1.5 | Aesthetic score: 5.78 -> 5.93 | Improved |
| use Union-ControlNet instead of Dual-ControlNet (ablation) | Composer · SD-v1.5 | step/s: 4.16 -> 11.69 | Improved |
| use Union-ControlNet instead of Dual-ControlNet (ablation) | Composer · SD-v1.5 | cycle consistency L_struct: 35.9 -> 36.8 | Hurt |
| use Union-ControlNet-XL instead of Dual-ControlNet on SDXL (ablation) | Composer-XL · SDXL | step/s: 0.90 -> 2.65 | Improved |
| use Union-ControlNet-XL instead of Dual-ControlNet on SDXL (ablation) | Composer-XL · SDXL | ImageReward: 1.12 -> 1.17 | Improved |
| add composition control module to SD-v15 (explicit planning) | SD-v15 · MS-COCO | Aesthetic score: 5.52 -> 5.89 | Improved |
| add composition control module to SD-v15 (explicit planning) | SD-v15 · MS-COCO | FID: 9.88 -> 11.04 | Hurt |
| remove LVLM planning stage (reference-free variant) | Composer · MS-COCO | runtime seconds: 7.7 -> 4.6 | Improved |
| increase aesthetic training images beyond 10K (ablation) | Composer · MS-COCO | CLIP Score | Improved |
| switch backbone from SD-v1.5 to SDXL | Composer · Parti | Aesthetic score: 5.93 -> 5.99 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- RealCompo — RealCompo: Balancing Realism and Compositionality Improves Text-to-Image Diffusion Models (NeurIPS 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Can Machines Understand Composition? Dataset and Benchmark for Photographic Image Composition Embedding and Understanding (CVPR 2025 · similarity 0.87)
- IterComp: Iterative Composition-Aware Feedback Learning from Model Gallery for Text-to-Image Generation (ICLR 2025 · similarity 0.86)
- Progressive Compositionality in Text-to-Image Generative Models (ICLR 2025 · similarity 0.86)
- Bridging Cognitive Gap: Hierarchical Description Learning for Artistic Image Aesthetics Assessment (AAAI 2026 · similarity 0.86)
- ArtBank: Artistic Style Transfer with Pre-trained Diffusion Model and Implicit Style Prompt Bank (AAAI 2024 · similarity 0.85)
- Stylebreeder: Exploring and Democratizing Artistic Styles through Text-to-Image Models (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.
