Research map / Image generation / Editing and style transfer
BigColor: Colorization Using a Generative Color Prior for Natural Images
Geonung Kim, Kyoungkook Kang, Seongtae Kim, Hwayoon Lee, Sehoon Kim, Jonghyun Kim, Seung-Hwan Baek, Sunghyun Cho
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 color augmentation to discriminator real images (ablation) | BigColor · ImageNet-1K val | colorfulness: 36.157 -> 40.006 | Improved |
| add color augmentation to discriminator real images (ablation) | BigColor · ImageNet-1K val | FID: 1.288 -> 1.243 | Improved |
| also augment the generated image (ablation) | BigColor · ImageNet-1K val | FID: 1.243 -> 1.604 | Hurt |
| initialize G and D from pretrained BigGAN (ablation) | BigColor · ImageNet 100-class subset | FID: 7.201 -> 5.714 | Improved |
| initialize only G from pretrained BigGAN (ablation) | BigColor · ImageNet 100-class subset | FID: 7.201 -> 8.058 | Mixed |
| add adversarial loss (ablation) | BigColor · ImageNet 100-class subset | FID: 7.692 -> 5.714 | Improved |
| add adversarial loss (ablation) | BigColor · ImageNet 100-class subset | classification accuracy: 75.52 -> 81.44 | Improved |
| add residual learning to encoder (ablation) | BigColor · ImageNet 100-class subset | FID: 6.523 -> 5.980 | Improved |
| add class-conditioned batch normalization to encoder (ablation) | BigColor · ImageNet 100-class subset | FID: 5.980 -> 5.714 | Improved |
| add plain batch normalization without residual learning (ablation) | BigColor encoder · ImageNet 100-class subset | FID: 6.523 -> 7.286 | Hurt |
| increase spatial feature resolution to 64x64 (ablation) | BigColor · ImageNet | FID | Mixed |
| use BigColor instead of ToVivid on complex images | BigColor · ImageNet-1K val (100 complex images) | classification accuracy: 65.000 -> 74.000 | Improved |
Later papers that compared with this method
2 later accepted papers report BigColor as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Control Color: Multimodal Diffusion-Based Interactive Image Colorization (IJCV 2025 · better on 2 of 2 shared results)
- DDColor: Towards Photo-Realistic Image Colorization via Dual Decoders (ICCV 2023 · better on 2 of 2 shared results)
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
- Towards Vivid and Diverse Image Colorization With Generative Color Prior (ICCV 2021 · similarity 0.91)
- DDColor: Towards Photo-Realistic Image Colorization via Dual Decoders (ICCV 2023 · similarity 0.89)
- Colorful Image Colorization (ECCV 2016 · similarity 0.89)
- Learning Diverse Image Colorization (CVPR 2017 · similarity 0.89)
- GenColor: Generative and Expressive Color Enhancement with Pixel-Perfect Texture Preservation (NeurIPS 2025 · similarity 0.88)
- BiGR: Harnessing Binary Latent Codes for Image Generation and Improved Visual Representation Capabilities (ICLR 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.
