Research map / Image generation / Text-to-image diffusion
Beyond Isolated Words: Diffusion Brush for Handwritten Text-Line Generation
Gang Dai, Yifan Zhang, Yutao Qin, Qiangya Guo, Shuangping Huang, Shuicheng Yan
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 column- and row-wise masked style module (vs no masking) (ablation) | DiffBrush · IAM | HWD: 1.82 -> 1.47 | Improved |
| use column/row-wise masking instead of random masking (ablation) | DiffBrush · IAM | HWD: 1.75 -> 1.47 | Improved |
| add line and word content discriminators (ablation) | DiffBrush · IAM | D_CER | Improved |
| use content discriminators instead of CTC recognizer (ablation) | DiffBrush · IAM | HWD: 1.67 -> 1.41 | Improved |
| use content discriminators instead of CTC recognizer (ablation) | DiffBrush · IAM | D_CER: 16.53 -> 8.59 | Improved |
| use 3D CNN line discriminator instead of 2D (ablation) | DiffBrush · IAM | D_CER: 13.18 -> 8.59 | Improved |
| switch word discriminator from 2D to 3D CNN (ablation) | DiffBrush · IAM | D_CER: 8.59 -> 9.58 | No clear effect |
| set masking ratio to 0.5 instead of 0 (ablation) | DiffBrush w/o discriminators · IAM | HWD: 1.72 -> 1.47 | Improved |
| set masking ratio to 0.75 instead of 0.5 (ablation) | DiffBrush w/o discriminators · IAM | HWD: 1.47 -> 1.58 | Hurt |
| generate whole text-lines directly instead of assembling One-DM words | DiffBrush · IAM | HWD: 2.17 -> 1.41 | Improved |
| set content loss weight lambda to 0.05 instead of 0.1 (ablation) | DiffBrush · IAM | D_CER: 9.87 -> 8.59 | Improved |
| add DiffBrush synthetic data to recognizer training | OCR recognizer · IAM test | relative improvement: 20.07% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VATr — VATr++: Choose Your Words Wisely for Handwritten Text Generation (TPAMI 2024) · better on 4 of 4 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
- Brush Your Text: Synthesize Any Scene Text on Images via Diffusion Model (AAAI 2024 · similarity 0.87)
- Layout-Agnostic Scene Text Image Synthesis with Diffusion Models (CVPR 2024 · similarity 0.87)
- Text to Sketch Generation with Multi-Styles (NeurIPS 2025 · similarity 0.87)
- FontDiffuser: One-Shot Font Generation via Denoising Diffusion with Multi-Scale Content Aggregation and Style Contrastive Learning (AAAI 2024 · similarity 0.87)
- Diff-Aid: Inference-time Adaptive Interaction Denoising for Rectified Text-to-Image Generation (NeurIPS 2026 · similarity 0.86)
- Decoupling Layout from Glyph in Online Chinese Handwriting Generation (ICLR 2025 · similarity 0.86)
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.
