Research map / Image generation / Editing and style transfer
DiffDecompose: Layer-Wise Decomposition of Alpha-Composited Images via Diffusion Transformers
Zitong Wang, Hang Zhao, Qianyu Zhou, Xuequan Lu, Xiangtai Li, Hao Yang, Bo Yang, Yiren Song
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 |
|---|---|---|---|
| fine-tune AlphaVAE on AlphaBlend (ablation) | DiffDecompose · AlphaBlend Translucent Flare (foreground) | RMSE: 46.2439 -> 21.2585 | Improved |
| fine-tune AlphaVAE on AlphaBlend (ablation) | DiffDecompose · AlphaBlend X-ray Contraband (foreground) | RMSE: 7.0811 -> 6.6880 | Improved |
| fine-tune AlphaVAE on AlphaBlend (ablation) | DiffDecompose · AlphaBlend Glassware (foreground) | LPIPS: 0.0579 -> 0.0578 | No clear effect |
| add Layer Position Encoding Cloning (LPEC) (ablation) | DiffDecompose · AlphaBlend X-ray Contraband | RMSE: 16.5379 -> 3.8903 | Improved |
| add LPEC (ablation) | DiffDecompose · AlphaBlend Glassware | RMSE: 24.6044 -> 7.9938 | Improved |
| add LPEC (ablation) | DiffDecompose · AlphaBlend Translucent Flare | RMSE: 32.4708 -> 16.4373 | Improved |
| add LPEC (ablation) | DiffDecompose · AlphaBlend Translucent Occlusion | RMSE: 21.2568 -> 18.9561 | Improved |
| add In-Context Decomposition (ICD) (ablation) | inpainting-style baseline · AlphaBlend subtasks | qualitative layer output | Improved |
| replace mask-based inpainting baselines with DiffDecompose | LOGO-L · LOGO-L | FID: 23.71 -> 24.20 | Hurt |
| replace mask-based inpainting baselines with DiffDecompose | AlphaBlend Translucent Flare Removal · AlphaBlend Translucent Flare Removal | RMSE: 63.1518 -> 16.4373 | Improved |
| replace Inpaint Anything with DiffDecompose | AlphaBlend Glassware Decomposition · AlphaBlend Glassware Decomposition | RMSE: 8.2335 -> 7.9938 | Mixed |
Closest papers by meaning
- PixelART: Image-to-Layer Decomposition without Latents or Text-to-Image Pretraining (NeurIPS 2026 · similarity 0.88)
- Diff-Shadow: Global-guided Diffusion Model for Shadow Removal (AAAI 2025 · similarity 0.87)
- DRIP: Unleashing Diffusion Priors for Joint Foreground and Alpha Prediction in Image Matting (NeurIPS 2024 · similarity 0.86)
- Compositional Image Decomposition with Diffusion Models (ICML 2024 · similarity 0.86)
- Generative Image Layer Decomposition with Visual Effects (CVPR 2025 · similarity 0.86)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.85)
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.
