Research map / Image generation / Autoregressive and latent generation
DeepVecFont-v2: Exploiting Transformers To Synthesize Vector Fonts With Higher Quality
Yuqing Wang, Yizhi Wang, Longhui Yu, Yuesheng Zhu, Zhouhui Lian
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 |
|---|---|---|---|
| replace DeepVecFont with DeepVecFont-v2 | English vector font synthesis · English testing dataset | reconstruction error: 0.056 -> 0.052 | Improved |
| replace DeepVecFont with DeepVecFont-v2 | Chinese vector font synthesis · Chinese testing dataset | reconstruction error: 0.086 -> 0.080 | Improved |
| add relaxation representation (ablation) | Transformer-based base model · English testing dataset | reconstruction error: 0.0588 -> 0.0557 | Improved |
| add Bézier curve alignment loss (ablation) | Base model with relaxation representation · English testing dataset | reconstruction error: 0.0557 -> 0.0529 | Improved |
| add self-refinement (ablation) | Model with relaxation representation and Bézier curve alignment loss · English testing dataset | reconstruction error: 0.0529 -> 0.0519 | Improved |
| increase Bézier-curve sampling points from 0 to 1 (ablation) | DeepVecFont-v2 · English testing dataset | reconstruction error: 0.0557 -> 0.0543 | Improved |
| increase Bézier-curve sampling points from 1 to 3 (ablation) | DeepVecFont-v2 · English testing dataset | reconstruction error: 0.0543 -> 0.0529 | Improved |
| increase Bézier-curve sampling points from 3 to 6 (ablation) | DeepVecFont-v2 · English testing dataset | reconstruction error: 0.0529 -> 0.0526 | Improved |
| increase Bézier-curve sampling points from 6 to 9 (ablation) | DeepVecFont-v2 · English testing dataset | reconstruction error: 0.0526 -> 0.0524 | Improved |
| increase Bézier-curve sampling points from 9 to 12 (ablation) | DeepVecFont-v2 · English testing dataset | reconstruction error: 0.0524 -> 0.0520 | Improved |
Later papers that compared with this method
1 later accepted paper reports DeepVecFont-v2 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- VecFusion: Vector Font Generation with Diffusion (CVPR 2024 · better on 3 of 3 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
- VecGlypher: Unified Vector Glyph Generation with Language Models (CVPR 2026 · similarity 0.89)
- VecFusion: Vector Font Generation with Diffusion (CVPR 2024 · similarity 0.89)
- DeepCalliFont: Few-Shot Chinese Calligraphy Font Synthesis by Integrating Dual-Modality Generative Models (AAAI 2024 · similarity 0.86)
- VQ-FONT: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization (AAAI 2024 · similarity 0.85)
- StrokeNUWA—Tokenizing Strokes for Vector Graphic Synthesis (ICML 2024 · similarity 0.84)
- Vector Grimoire: Codebook-based Shape Generation under Raster Image Supervision (ICML 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.
