Research map / Image generation / Editing and style transfer
DACoN: DINO for Anime Paint Bucket Colorization with Any Number of Reference Images
Kazuma Nagata, Naoshi Kaneko
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 DINO-guided feature consistency loss (ablation) | DACoN 1-shot · Keyframe full test set | Acc: 63.95 -> 67.87 | Improved |
| add DINO-guided feature consistency loss (ablation) | DACoN 5-shot · Keyframe full test set | Acc: 72.57 -> 73.25 | Improved |
| add feature consistency loss to SD-based variant (ablation) | DACoN (SD) · Keyframe 1-shot | Acc: 60.96 -> 62.26 | Improved |
| keep colored lines instead of unifying to black (ablation) | DACoN 1-shot · Keyframe full test set | Acc: 64.32 -> 67.87 | Improved |
| increase reference images from 1 to 5 | DACoN · Keyframe full test set | Acc: 67.87 -> 73.25 | Improved |
| use all color design sheet images instead of 5 | DACoN · Keyframe full test set | Acc: 73.25 -> 74.31 | Improved |
| increase reference images from 1 to 5 | BasicPBC-Ref · Keyframe 8-character subset | Acc-Thresh: 63.46 -> 64.59 | No clear effect |
| add following frame as reference | DACoN · Consecutive frame 3D rendered | Acc: 84.71 -> 88.93 | Improved |
| add following frame as reference | DACoN · Consecutive frame hand-drawn | Acc: 87.02 -> 91.32 | Improved |
| add 5-shot color design sheet to preceding frame | DACoN · Consecutive frame 3D rendered | Acc: 84.71 -> 85.61 | Improved |
| add 5-shot color design sheet to preceding and following frames | DACoN · Consecutive frame 3D rendered | Acc: 88.93 -> 88.83 | No clear effect |
| replace concat+MLP fusion with cross-attention (ablation) | DACoN 1-shot · Keyframe 3D rendered | Acc: 67.87 -> 47.60 | Hurt |
Closest papers by meaning
- MangaNinja: Line Art Colorization with Precise Reference Following (CVPR 2025 · similarity 0.86)
- Towards High-resolution and Disentangled Reference-based Sketch Colorization (CVPR 2026 · similarity 0.86)
- Learning Inclusion Matching for Animation Paint Bucket Colorization (CVPR 2024 · similarity 0.86)
- Image Referenced Sketch Colorization Based on Animation Creation Workflow (CVPR 2025 · similarity 0.84)
- AniDoc: Animation Creation Made Easier (CVPR 2025 · similarity 0.83)
- SSIMBaD: Sigma Scaling with SSIM-Guided Balanced Diffusion for AnimeFace Colorization (NeurIPS 2025 · similarity 0.82)
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.
