Research map / Image generation / GANs
HistoGAN: Controlling Colors of GAN-Generated and Real Images via Color Histograms
Mahmoud Afifi, Marcus A. Brubaker, Michael S. Brown
Check 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 |
|---|---|---|---|
| condition generated images on target color histograms instead of relying on StyleGAN latent projection | HistoGAN vs. StyleGAN · Faces, Flowers, Cats, Dogs, Birds, Anime, Hands, Landscape, Bedrooms, Cars, Aerial Scenes | RGB-uv histogram Hellinger distance: 0.6724 -> 0.3088; 0.7038 -> 0.2957; 0.5303 -> 0.2639; 0.5658 -> 0.3058; 0.6091 -> 0.3064; 0.5929 -> 0.1798; 0.3709 -> 0. | Improved |
| set histogram-loss scale alpha to 0.2 rather than 2 (ablation) | HistoGAN · Hands | FID: 2.2438 -> 1.9950 | Improved |
| set histogram-loss scale alpha to 4 rather than 2 (ablation) | HistoGAN · Hands | FID: 2.2438 -> 6.8750 | Hurt |
| set histogram-loss scale alpha to 8 rather than 2 (ablation) | HistoGAN · Hands | FID: 2.2438 -> 9.4101 | Hurt |
| set histogram-loss scale alpha to 16 rather than 2 (ablation) | HistoGAN · Hands | FID: 2.2438 -> 15.747 | Hurt |
| replace the Gaussian histogram kernel with an inverse-quadratic kernel (ablation) | HistoGAN · GAN image generation | training stability | Improved |
| add the variance loss to discourage global color shifts (ablation) | ReHistoGAN · real-image recoloring | global color-cast artifacts | Improved |
| add encoder-decoder skip connections and end-to-end fine-tuning of the HistoGAN head (ablation) | ReHistoGAN · real-image recoloring | preservation of fine details | Improved |
| use a Laplacian rather than Sobel kernels for the reconstruction loss (ablation) | ReHistoGAN · real-image recoloring | qualitative result quality | Improved |
| increase the discriminator-loss weight so the model prioritizes realism (ablation) | ReHistoGAN · real-image recoloring | target-color matching | Hurt |
Closest papers by meaning
- PalGAN: Image Colorization with Palette Generative Adversarial Networks (ECCV 2022 · similarity 0.89)
- Towards Vivid and Diverse Image Colorization With Generative Color Prior (ICCV 2021 · similarity 0.88)
- Differentiable Histogram Loss Functions for Intensity-based Image-to-Image Translation (TPAMI 2023 · similarity 0.87)
- CartoonGAN: Generative Adversarial Networks for Photo Cartoonization (CVPR 2018 · similarity 0.87)
- Learning Representations for Automatic Colorization (ECCV 2016 · similarity 0.86)
- Secret Lies in Color: Enhancing AI-Generated Images Detection with Color Distribution Analysis (CVPR 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.
