Research map / Image generation / Text-to-image diffusion
DreamLayer: Simultaneous Multi-Layer Generation via Diffusion Model
Junjia Huang, Pengxiang Yan, Jinhang Cai, Jiyang Liu, Zhao Wang, Yitong Wang, Xinglong Wu, Guanbin 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 CACA (ablation) | DreamLayer (LSSA only) · multi-layer average | AES: 6.471 -> 6.625 | Improved |
| add LSSA (ablation) | DreamLayer (CACA only) · multi-layer average | CLIP: 34.004 -> 35.275 | Improved |
| add CACA (ablation) | baseline without LSSA · multi-layer average | FID: 57.241 -> 55.788 | Improved |
| add LSSA (ablation) | baseline without CACA · multi-layer average | CLIP: 33.808 -> 33.727 | No clear effect |
| add LSSA (ablation) | baseline without CACA · multi-layer average | FID: 57.241 -> 56.598 | Improved |
| add Information Retained Harmonization (ablation) | DreamLayer · two-layer test | AES: 6.967 -> 7.013 | Improved |
| add Information Retained Harmonization (ablation) | DreamLayer · four-layer test | FID: 57.073 -> 53.598 | Improved |
| add Information Retained Harmonization (ablation) | DreamLayer · two-layer test | CLIP: 34.587 -> 34.835 | Improved |
| increase T_H to 800 (T_H'=600) in IRH (ablation) | DreamLayer IRH · avg AES | AES: 6.625 -> 6.615 | Hurt |
| apply IRH late (T_H' below 600 near end of denoising) (ablation) | DreamLayer IRH · avg AES | AES: 6.625 -> 6.553 | Hurt |
| use DreamLayer instead of LayerDiffusion | multi-layer generation · four-layer test | FID: 79.997 -> 53.598 | Improved |
| use DreamLayer instead of SD v1.5 full-image generation | composite image · two-layer test | AES: 6.930 -> 7.013 | Improved |
Closest papers by meaning
- Masked Region Transformer for Layered Image Generation and Editing at Scale (CVPR 2026 · similarity 0.90)
- R&B: Region and Boundary Aware Zero-shot Grounded Text-to-image Generation (ICLR 2024 · similarity 0.87)
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 2024 · similarity 0.87)
- ART: Anonymous Region Transformer for Variable Multi-Layer Transparent Image Generation (CVPR 2025 · similarity 0.86)
- LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration (NeurIPS 2025 · similarity 0.86)
- Generative Image Layer Decomposition with Visual Effects (CVPR 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.
