Research map / Image generation / Text-to-image diffusion
CreatiLayout: Siamese Multimodal Diffusion Transformer for Creative Layout-to-Image Generation
Hui Zhang, Dexiang Hong, Yitong Wang, Jie Shao, Xinglong Wu, Zuxuan Wu, Yu-Gang Jiang
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 layout guidance with a cross-attention layout adapter (ablation) | SD3 · LayoutSAM-Eval | Spatial: 78.70 -> 88.43 | Improved |
| add layout as joint modality in M3-Attention (ablation) | SD3 · LayoutSAM-Eval | Spatial: 78.70 -> 79.15 | No clear effect |
| decouple M3-Attention into parallel image-text and image-layout branches (SiamLayout) (ablation) | SD3 with M3-Attention · LayoutSAM-Eval | Spatial: 79.15 -> 92.67 | Improved |
| use SiamLayout instead of layout adapter (ablation) | SD3 with layout adapter · LayoutSAM-Eval | Spatial: 88.43 -> 92.67 | Improved |
| use SiamLayout instead of layout adapter (ablation) | SD3 with layout adapter · LayoutSAM-Eval | Texture: 73.56 -> 77.21 | Improved |
| add layout guidance to SD3 on T2I-CompBench via LLM planned layouts | SD3 · T2I-CompBench | Spatial: 32.00 -> 47.36 | Improved |
| add layout guidance to SD3 | SD3 · T2I-CompBench | Shape: 58.85 -> 61.32 | Improved |
| add region-aware loss (ablation) | SiamLayout training · LayoutSAM | convergence speed | Improved |
| add biased time step sampling toward larger steps (ablation) | SiamLayout training · LayoutSAM | convergence speed | Improved |
| fine-tune Llama3.1 into LayoutDesigner (SFT) for caption-to-layout | Llama3.1-8b-it · Layout planning (global caption) | accuracy: +49.89 | Improved |
| fine-tune Llama3.1 into LayoutDesigner (SFT) for layout optimization | Llama3.1-8b-it · Layout planning (center point) | accuracy: +26.65 | Improved |
| use SiamLayout instead of InstanceDiffusion on coarse COCO | layout-to-image · COCO 2017 | AP: 51.9 -> 47.8 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- HiCo — HiCo: Hierarchical Controllable Diffusion Model for Layout-to-image Generation (NeurIPS 2024) · better on 2 of 2 shared results
- MIGC — MIGC: Multi-Instance Generation Controller for Text-to-Image Synthesis (CVPR 2024) · better on 1 of 1 shared result
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
- Rethinking The Training And Evaluation of Rich-Context Layout-to-Image Generation (NeurIPS 2024 · similarity 0.90)
- InstanceAssemble: Layout-Aware Image Generation via Instance Assembling Attention (NeurIPS 2025 · similarity 0.89)
- Build-A-Scene: Interactive 3D Layout Control for Diffusion-Based Image Generation (ICLR 2025 · similarity 0.89)
- HiCo: Hierarchical Controllable Diffusion Model for Layout-to-image Generation (NeurIPS 2024 · similarity 0.89)
- Towards Aligned Layout Generation via Diffusion Model with Aesthetic Constraints (ICLR 2024 · similarity 0.88)
- OverLayBench: A Benchmark for Layout-to-Image Generation with Dense Overlaps (NeurIPS 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.
