Research map / Image generation / Image restoration
DDColor: Towards Photo-Realistic Image Colorization via Dual Decoders
Xiaoyang Kang, Tao Yang, Wenqi Ouyang, Peiran Ren, Lingzhi Li, Xuansong Xie
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 the query-based color decoder (ablation) | DDColor · ImageNet (val5k) | FID: 5.93 -> 3.92 | Improved |
| add colorfulness loss (ablation) | DDColor · ImageNet (val5k) | Delta CF: 2.52 -> 0.05 | Improved |
| add colorfulness loss (ablation) | DDColor · ImageNet (val5k) | FID: 4.01 -> 3.92 | Improved |
| use multi-scale (3 scales) features in color decoder instead of single scale (ablation) | DDColor · ImageNet (val5k) | FID: 4.44 -> 3.92 | Improved |
| cross-attention before self-attention in color decoder block (ablation) | DDColor · ImageNet (val5k) | FID: 3.98 -> 3.92 | Improved |
| replace cross-attention by a second self-attention layer (ablation) | DDColor · ImageNet (val5k) | FID: 3.92 -> 8.74 | Hurt |
| increase number of color queries from 100 to 200 (ablation) | DDColor · ImageNet (val5k) | FID: 3.92 -> 3.96 | No clear effect |
| reduce color queries from 100 to 20 (ablation) | DDColor · ImageNet (val5k) | FID: 3.92 -> 4.02 | Hurt |
| use larger ConvNeXt backbone (T to L) (ablation) | DDColor · ImageNet (val5k) | FID: 4.38 -> 3.92 | Improved |
| DDColor versus baselines on colorfulness score (CF) | DDColor-large · ImageNet (val5k) | CF: 38.26 vs BigColor 39.74 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ColorFormer — ColorFormer: Image Colorization via Color Memory Assisted Hybrid-Attention Transformer (ECCV 2022) · better on 5 of 5 shared results
- BigColor — BigColor: Colorization Using a Generative Color Prior for Natural Images (ECCV 2022) · 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
- Colorful Image Colorization (ECCV 2016 · similarity 0.92)
- ColorFormer: Image Colorization via Color Memory Assisted Hybrid-Attention Transformer (ECCV 2022 · similarity 0.91)
- Learning a Deep Convolutional Network for Colorization in Monochrome-Color Dual-Lens System (AAAI 2019 · similarity 0.91)
- Pixelated Semantic Colorization (IJCV 2019 · similarity 0.90)
- Enabling Region-Specific Control via Lassos in Point-Based Colorization (AAAI 2025 · similarity 0.88)
- Color-Oriented Redundancy Reduction in Dataset Distillation (NeurIPS 2024 · similarity 0.88)
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.
