Research map / Image generation / Editing and style transfer
GAN-Based Unpaired Chinese Character Image Translation via Skeleton Transformation and Stroke Rendering
Read the paperCheck 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 |
|---|---|---|---|
| use three-stage skeleton transformation plus stroke rendering instead of direct character-to-character TNet (ablation) | ChiroGAN · StdFont-4 | MS ACC (%): 94.15 -> 99.72 | Helped |
| three-stage architecture vs Exp-TNet (ablation) | ChiroGAN · StdFont-4 | content error rate (%): 0.88 -> 0.32 | Helped |
| three-stage architecture vs Exp-TNet (ablation) | ChiroGAN · StdFont-4 | style error rate (%): 1.18 -> 0.14 | Helped |
| three-stage architecture vs Exp-TNet (ablation) | ChiroGAN · StdFont-4 user study | average score: 0.82 -> 2.13 | Helped |
| replace multi-domain StarGAN with ChiroGAN (ablation) | ChiroGAN · StdFont-4 | style error rate (%): 3.36 -> 0.14 | Helped |
| add joint training of TNet with pre-trained RNet (Stage-II) (ablation) | ChiroGAN · StdFont-4 | skeleton quality (qualitative) | Helped |
| add pre-training of TNet (ablation) | ChiroGAN · StdFont-4 | skeleton details (qualitative) | Helped |
| add content loss L_cont on encoder features (ablation) | ChiroGAN · handwriting input | content consistency (qualitative) | Helped |
| replace morphological skeletonization with convolutional ENet (ablation) | ENet · skeleton extraction | time per image: 10 seconds -> 0.02 seconds | Helped |
| ChiroGAN vs supervised zi2zi | ChiroGAN · StdFont-4 user study test-I | average score: 2.84 vs 2.13 | Hurt |
| ChiroGAN vs CycleGAN on dissimilar Clerical2YouYuan | ChiroGAN · StdFont-4 C2Y | IOU: 0.124 -> 0.346 | Helped |
Closest papers by meaning
- StrokeGAN: Reducing Mode Collapse in Chinese Font Generation via Stroke Encoding (AAAI 2021 · similarity 0.87)
- MangaGAN: Unpaired Photo-to-Manga Translation Based on The Methodology of Manga Drawing (AAAI 2021 · similarity 0.87)
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.86)
- FontRL: Chinese Font Synthesis via Deep Reinforcement Learning (AAAI 2021 · similarity 0.86)
- DeepCalliFont: Few-Shot Chinese Calligraphy Font Synthesis by Integrating Dual-Modality Generative Models (AAAI 2024 · similarity 0.85)
- A Multi-Style Chinese Characters Writing Intelligent Tool Based on Small-scale Training Data (AAAI 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.