Research map / Image generation / Editing and style transfer
Streamlining Image Editing with Layered Diffusion Brushes
Peyman Gholami, Robert Xiao
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 |
|---|---|---|---|
| use mid-range regeneration step r (r=12) instead of very large r (r=23) (ablation) | LDB · regeneration step ablation | HPS: 0.33 -> 0.29 | Improved |
| avoid very small regeneration step r=2 (ablation) | LDB · regeneration step ablation | LPIPS / PSNR: LPIPS 0.04, PSNR 27.03 | Hurt |
| use large regeneration step r (near N) (ablation) | LDB · regeneration step ablation | CS-L: 22.98 | Hurt |
| delay blending of cached latent (b=N-2 rather than b=n) (ablation) | LDB · blending step ablation | LPIPS / PSNR: LPIPS 0.17, PSNR 11.61 at b=n | Improved |
| increase blending step b toward N-1 (ablation) | LDB · blending step ablation | CS-D: 0.32 at b = N-1 | Improved |
| increase blending step b (ablation) | LDB · blending step ablation | CS-L | No clear effect |
| reduce editing steps to n=4 (ablation) | LDB · speed analysis | HPS / CS-D / latency: HPS 0.34, CS-D 0.35, 140ms | Mixed |
| add latent caching for edits | LDB · 512x512 editing | time per edit: 53x faster than BrushNet; 140 ms per edit | Improved |
| use LDB instead of IP2P and SDI | LDB · user study | SUS score: 38.21% -> 80.35% | Improved |
| compare LDB against BrushNet on PIE-Bench | LDB · PIE-Bench | IR: 72.77 -> 86.02 | Improved |
| compare LDB against BrushNet on MagicBrush text alignment | LDB · MagicBrush | CS-D: 12.92 -> 9.54 | Hurt |
Closest papers by meaning
- Stable Flow: Vital Layers for Training-Free Image Editing (CVPR 2025 · similarity 0.90)
- Move Anything with Layered Scene Diffusion (CVPR 2024 · similarity 0.89)
- Schedule Your Edit: A Simple yet Effective Diffusion Noise Schedule for Image Editing (NeurIPS 2024 · similarity 0.89)
- DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework (AAAI 2025 · similarity 0.89)
- h-Edit: Effective and Flexible Diffusion-Based Editing via Doob's h-Transform (CVPR 2025 · similarity 0.89)
- High-Fidelity Diffusion-Based Image Editing (AAAI 2024 · similarity 0.89)
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.
