Research map / Image generation / Editing and style transfer
Reference-Based Sketch Image Colorization Using Augmented-Self Reference and Dense Semantic Correspondence
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 similarity-based triplet loss (ablation) | Reference-based sketch colorization model · ImageNet Cat, Dog, Car; CelebA; Tag2pix; Yumi's Cells; Edges→Shoes | FID: 77.39 -> 74.12; 109.49 -> 102.83; 54.07 -> 52.23; 53.58 -> 47.15; 47.68 -> 45.34; 51.34 -> 49.29; 79.85 -> 78.32 | Helped |
| replace SCFT with simple addition aggregation (ablation) | Reference-based sketch colorization model · ImageNet Cat, Dog; CelebA; Tag2pix; Yumi's Cells; Edges2Shoes | FID: 74.12 -> 78.47; 102.83 -> 103.73; 52.23 -> 55.80; 47.15 -> 51.94; 45.34 -> 47.72; 49.29 -> 47.67; 78.32 -> 117.15 | Mixed |
| replace SCFT with AdaIN aggregation (ablation) | Reference-based sketch colorization model · ImageNet Cat, Dog, Car; CelebA; Tag2pix; Yumi's Cells; Edges2Shoes | FID: 74.12 -> 75.17; 102.83 -> 105.72; 52.23 -> 52.85; 47.15 -> 50.61; 45.34 -> 52.81; 49.29 -> 45.36; 78.32 -> 88.46 | Mixed |
| add adversarial loss (ablation) | Reference-based sketch colorization model · ImageNet Cat, Dog, Car; CelebA; Tag2pix; Yumi's Cells; Edges→Shoes | FID: 82.10 -> 78.56; 143.76 -> 110.86; 68.45 -> 56.54; 77.70 -> 54.75; 58.00 -> 48.71; 52.86 -> 51.96; 91.10 -> 82.55 | Helped |
| add perceptual and style losses (ablation) | Reference-based sketch colorization model with reconstruction and adversarial losses · ImageNet Cat, Dog, Car; CelebA; Tag2pix; Yumi's Cells; Edges→Shoes | FID: 78.56 -> 77.39; 110.86 -> 109.49; 56.54 -> 54.07; 54.75 -> 53.58; 48.71 -> 47.68; 51.96 -> 51.34; 82.55 -> 79.85 | Helped |
| add similarity-based triplet loss (ablation) | Model trained with reconstruction, adversarial, perceptual, and style losses · ImageNet Cat, Dog, Car; CelebA; Tag2pix; Yumi's Cells; Edges→Shoes | FID: 77.39 -> 74.12; 109.49 -> 102.83; 54.07 -> 52.23; 53.58 -> 47.15; 47.68 -> 45.34; 51.34 -> 49.29; 79.85 -> 78.32 | Helped |
| add similarity-based triplet loss supervision (ablation) | Reference-based sketch colorization model · SPair-71k Cat, Dog, Car | SC-PSNR (dB): 12.76 -> 13.23; 13.73 -> 14.37; 10.56 -> 11.34 | Helped |
Closest papers by meaning
- Image Referenced Sketch Colorization Based on Animation Creation Workflow (CVPR 2025 · similarity 0.91)
- MangaNinja: Line Art Colorization with Precise Reference Following (CVPR 2025 · similarity 0.88)
- Automatic Controllable Colorization via Imagination (CVPR 2024 · similarity 0.87)
- One-Shot Reference-based Structure-Aware Image to Sketch Synthesis (AAAI 2025 · similarity 0.86)
- Learning Inclusion Matching for Animation Paint Bucket Colorization (CVPR 2024 · similarity 0.84)
- Text to Sketch Generation with Multi-Styles (NeurIPS 2025 · similarity 0.84)
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.