Research map / Image generation / Text-to-image diffusion
LayerDiff: Exploring Text-guided Multi-layered Composable Image Synthesis via Layer-Collaborative Diffusion Model
Runhui Huang, Kaixin Cai, Jianhua Han, Xiaodan Liang, Renjing Pei, Guansong Lu, Songcen Xu, Wei Zhang, Hang Xu
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 self-mask guidance (SMG) sampling (ablation) | LayerDiff · MLCID test, two layers | FID: 35.1 -> 21.3 | Improved |
| add self-mask guidance (SMG) sampling (ablation) | LayerDiff · MLCID test, four layers | FID: 83.9 -> 58.7 | Improved |
| add text-guided intra-layer attention (ablation) | LayerDiff · MLCID ablation subset, two layers | FID: 358.0 -> 45.8 | Improved |
| add mask dilation (kernel 5) (ablation) | LayerDiff · MLCID ablation subset, two layers | FID: 53.7 -> 45.8 | Improved |
| add layer-specific prompt enhancer (self+cross attention) (ablation) | LayerDiff · MLCID ablation subset, three layers | FID: 92.2 -> 92.9 | No clear effect |
| add layer-specific prompt enhancer (self+cross attention) (ablation) | LayerDiff · MLCID ablation subset, two layers | FID: 45.8 -> 46.7 | Hurt |
| replace FC mask information injection with zero-initialized scalar (ablation) | LayerDiff · MLCID ablation subset, four layers | FID: 140.2 -> 112.8 | Improved |
| replace FC mask information injection with zero-initialized scalar (ablation) | LayerDiff · MLCID ablation subset, two layers | FID: 46.7 -> 50.1 | Hurt |
| use SMG instead of SAG sampling guidance (ablation) | LayerDiff · MLCID 1k prompts, all layers | FID: 92.6 -> 88.9 | Improved |
| generate three and four layers with limited training data | LayerDiff · MLCID test, three and four layers | FID: SD 25.9 vs LayerDiff 58.9 (three layers) | Hurt |
Closest papers by meaning
- R&B: Region and Boundary Aware Zero-shot Grounded Text-to-image Generation (ICLR 2024 · similarity 0.88)
- Devil is in the Detail: Towards Injecting Fine Details of Image Prompt in Image Generation via Conflict-free Guidance and Stratified Attention (CVPR 2025 · similarity 0.88)
- Generative Image Layer Decomposition with Visual Effects (CVPR 2025 · similarity 0.88)
- LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration (NeurIPS 2025 · similarity 0.87)
- DesignDiffusion: High-Quality Text-to-Design Image Generation with Diffusion Models (CVPR 2025 · similarity 0.87)
- Prompt-Free Diffusion: Taking "Text" out of Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.87)
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.
