Research map / Image generation / Text-to-image diffusion
Referring Layer Decomposition
Fangyi Chen, Yaojie Shen, Lu Xu, Ye Yuan, Shu Zhang, Yulei Niu, Longyin Wen
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 |
|---|---|---|---|
| train on RefLade instead of MuLAn (50K layers) | RefLayer · RefLade test (foreground) | HPA: 0.3852 -> 0.4629 | Improved |
| train on RefLade instead of MuLAn (50K layers) | RefLayer · RefLade test (background) | HPA: 0.3459 -> 0.5932 | Improved |
| scale RefLade training from 50K to 1M layers | RefLayer · RefLade test (foreground) | HPA: 0.4629 -> 0.4685 | Improved |
| scale RefLade training from 50K to 1M layers | RefLayer · RefLade test (background) | HPA: 0.5932 -> 0.5587 | Hurt |
| fine-tune on high-quality RefLadeQ after RefLade pretraining | RefLayer · RefLade test (background) | HPA: 0.5587 -> 0.6682 | Improved |
| fine-tune on high-quality RefLadeQ after RefLade pretraining | RefLayer · RefLade test (foreground) | HPA: 0.4685 -> 0.4813 | Improved |
| use spatial mask prompt instead of text prompt (ablation) | RefLayer · RefLade test | HPA_frgd: 0.2403 -> 0.4842 | Improved |
| add text to mask prompt (ablation) | RefLayer · RefLade test | HPA_occ: 0.4270 -> 0.4403 | Improved |
| add text to mask prompt (ablation) | RefLayer · RefLade test | HPA_frgd: 0.4842 -> 0.4833 | Hurt |
| initialize from UltraEdit instead of vanilla SD3 (ablation) | RefLayer · RefLade test | HPA_frgd: 0.4275 -> 0.4698 | Improved |
| use checkerboard canvas instead of black canvas (ablation) | RefLayer (UltraEdit) · RefLade test | HPA_frgd: 0.4607 -> 0.4698 | Improved |
| use min-max normalization for HPA aggregation instead of sigmoid or log scaling (ablation) | HPA metric · human ELO agreement | correlation | Improved |
Closest papers by meaning
- Referring Image Editing: Object-level Image Editing via Referring Expressions (CVPR 2024 · similarity 0.88)
- Generative Image Layer Decomposition with Visual Effects (CVPR 2025 · similarity 0.88)
- PixelART: Image-to-Layer Decomposition without Latents or Text-to-Image Pretraining (NeurIPS 2026 · similarity 0.87)
- LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration (NeurIPS 2025 · similarity 0.87)
- MULAN: A Multi Layer Annotated Dataset for Controllable Text-to-Image Generation (CVPR 2024 · similarity 0.87)
- Improving Editability in Image Generation with Layer-wise Memory (CVPR 2025 · 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.
