Research map / Image generation / Editing and style transfer
AIComposer: Any Style and Content Image Composition via Feature Integration
Haowen Li, Zhenfeng Fan, Zhang Wen, Zhengzhou Zhu, Yunjin Li
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 CLIP image features (ImageCLIP) (ablation) | AIComposer · Extended benchmark (ablation) | CSD: 0.5083 -> 0.5284 | Improved |
| add initial latent blending (InitBlend) (ablation) | AIComposer · Extended benchmark (ablation) | CSD: 0.4671 -> 0.5284 | Improved |
| add initial latent blending (InitBlend) (ablation) | AIComposer · Extended benchmark (ablation) | LPIPS: 0.3376 -> 0.4195 | Hurt |
| add inversion of the blended latent (ablation) | AIComposer · Extended benchmark (ablation) | LPIPS: 0.5867 -> 0.4195 | Improved |
| add inversion of the blended latent (ablation) | AIComposer · Extended benchmark (ablation) | CLIP-T: 24.23 -> 28.75 | Improved |
| use fewer diffusion steps instead of full diffusion (ablation) | AIComposer · Extended benchmark (ablation) | Time: 19.67s -> 12.94s | Improved |
| use single-branch instead of dual-branch (ablation) | AIComposer · Extended benchmark (ablation) | LPIPS: 0.4945 -> 0.4195 | Improved |
| use single-branch instead of dual-branch (ablation) | AIComposer · Extended benchmark (ablation) | Time: 29.25s -> 12.94s | Improved |
| remove text prompt (null prompt) | AIComposer · Baseline benchmark (TF-ICON cross-domain) | CSD: 0.4948 -> 0.4743 | Hurt |
| remove text prompt (null prompt) | AIComposer · Extended benchmark (cross-domain) | CLIP-T: 29.29 -> 28.75 | Hurt |
| train MLP on 30K instead of 300 filtered triplets (ablation) | AIComposer MLP · Extended benchmark (MLP data ablation) | CLIP-T: 28.05 -> 28.75 | Improved |
| filter training triplets (300 CSGO) (ablation) | AIComposer MLP · Extended benchmark (MLP data ablation) | LPIPS: 0.4354 -> 0.4322 | Improved |
Closest papers by meaning
- ORCA: Hunting Compositional Failures in Text-to-Image Diffusion (NeurIPS 2026 · similarity 0.88)
- RealCompo: Balancing Realism and Compositionality Improves Text-to-Image Diffusion Models (NeurIPS 2024 · similarity 0.87)
- Conditional Balance: Improving Multi-Conditioning Trade-Offs in Image Generation (CVPR 2025 · similarity 0.86)
- Z*: Zero-shot Style Transfer via Attention Reweighting (CVPR 2024 · similarity 0.86)
- Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · similarity 0.86)
- CSGO: Content-Style Composition in Text-to-Image Generation (NeurIPS 2025 · similarity 0.86)
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.
