Research map / Image generation / Text-to-image diffusion
NumColor: Precise Numeric Color Control in Text-to-Image Generation
Muhammad Atif Butt, Diego Hernández, Alex Gomez-Villa, 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 NumColor (CTA + ColorBook) | FLUX · GenColorBench numeric ISCC-L1 | color accuracy (%): 13.82 -> 55.71 | Improved |
| add NumColor | FLUX · GenColorBench numeric CSS3/X11 | color accuracy (%): 5.86 -> 51.96 | Improved |
| transfer FLUX-trained ColorBook zero-shot | SD3 · GenColorBench numeric ISCC-L1 | color accuracy (%): 11.80 -> 49.23 | Improved |
| transfer FLUX-trained ColorBook zero-shot | PixArt-alpha · GenColorBench numeric ISCC-L1 | color accuracy (%): 9.45 -> 44.93 | Improved |
| add NumColor | FLUX · GenColorBench numeric (RGB) | color harmony (lower better): 247.4 -> 16.90 | Improved |
| inject color embeddings before T5 contextualization instead of after (ablation) | ColorBook · numeric color prompts | color-object binding | Improved |
| use k=1 hard nearest-anchor assignment (ablation) | ColorBook interpolation · numeric color prompts | visual quality | Hurt |
| increase softmax temperature tau (ablation) | ColorBook interpolation · numeric color prompts | color saturation | Hurt |
| apply NumColor to objects with strong canonical colors | FLUX w/ NumColor · GenColorBench ISCC-L1 by category | color accuracy (%): 62.1 (clothing) vs 36.2 (fruits & vegetables) | Mixed |
Closest papers by meaning
- Recolour What Matters: Region-Aware Colour Editing via Token-Level Diffusion (ECCV 2026 · similarity 0.90)
- GenColorBench: A Color Evaluation Benchmark for Text-to-Image Generation (CVPR 2026 · similarity 0.89)
- Control Color: Multimodal Diffusion-Based Interactive Image Colorization (IJCV 2025 · similarity 0.89)
- Color Alignment in Diffusion (CVPR 2025 · similarity 0.87)
- ColorConceptBench: A Benchmark for Probabilistic Color-Concept Understanding in Text-to-Image Models (NeurIPS 2026 · similarity 0.87)
- Free-Lunch Color-Texture Disentanglement for Stylized Image Generation (NeurIPS 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.
