Research map / Image generation / Text-to-image diffusion
ColorPeel: Color Prompt Learning with Diffusion Models via Color and Shape Disentanglement
Muhammad Atif Butt, Kai Wang, Javier Vazquez-Corral, Joost van de Weijer
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 cross-attention alignment (CAA) loss between color and shape tokens (ablation) | ColorPeel (3D) on Custom Diffusion · Color prompt learning (coarse-grained) | Delta E (CIE Lab): 48.47 -> 21.39 | Improved |
| add cross-attention alignment loss (ablation) | ColorPeel (3D) · Color prompt learning (coarse-grained) | MAE Hue (100% pixels): 78.43 -> 21.35 | Improved |
| increase CAA weight lambda from 0.2 to 1.0 (ablation) | ColorPeel (3D) · Color prompt learning (coarse-grained) | Delta E (CIE Lab): 21.39 -> 24.43 | Hurt |
| use 3D shapes instead of 2D shapes for color learning (ablation) | ColorPeel · Color prompt learning (coarse-grained) | Delta E (CIE Lab): 20.45 (2D) vs 21.39 (3D) Delta E; MAE Hue 100% 21.46 (2D) vs 21.35 (3D) | Mixed |
| learn color and shape tokens jointly from generated colored geometric shapes rather than naive personalization | Stable Diffusion v1.4 personalization · Color prompt learning (coarse-grained) | Delta E (CIE Lab): 48.47 (Custom Diffusion) -> 21.39 | Improved |
| specify exact RGB values via Rich-Text instead of color names in Stable Diffusion | Stable Diffusion v1.4 · Color prompt learning (coarse-grained) | Delta E (CIE Lab): 47.45 -> 36.62 | Improved |
| apply ColorPeel to DreamBooth | DreamBooth · Color prompt learning | color accuracy (qualitative) | Improved |
| replace plain full-color patches with colored shapes (color token learned from shapes) (ablation) | Custom Diffusion / personalization baselines · Color prompt learning | attention focus on color region (qualitative) | Improved |
Later papers that compared with this method
2 later accepted papers report ColorPeel as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Recolour What Matters: Region-Aware Colour Editing via Token-Level Diffusion (ECCV 2026 · better on 2 of 2 shared results)
- The Latent Color Subspace: Emergent Order in High-Dimensional Chaos (ICML 2026 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Free-Lunch Color-Texture Disentanglement for Stylized Image Generation (NeurIPS 2025 · similarity 0.88)
- Learning Generative Neural Networks for 3D Colorization (AAAI 2018 · similarity 0.87)
- ColorConceptBench: A Benchmark for Probabilistic Color-Concept Understanding in Text-to-Image Models (NeurIPS 2026 · similarity 0.86)
- DreamDistribution: Learning Prompt Distribution for Diverse In-distribution Generation (ICLR 2025 · similarity 0.85)
- Prompt-Free Diffusion: Taking "Text" out of Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.84)
- Minority-Focused Text-to-Image Generation via Prompt Optimization (CVPR 2025 · similarity 0.84)
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.
