Research map / Image generation / Text-to-image diffusion
DiffusionPen: Towards Controlling the Style of Handwritten Text Generation
Konstantina Nikolaidou, George Retsinas, Giorgos Sfikas, Marcus Liwicki
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 |
|---|---|---|---|
| use hybrid classification plus triplet style encoder instead of classification only (ablation) | DiffusionPen · IAM writer classification | accuracy (%): 68.04 -> 70.31 | Improved |
| use hybrid classification plus triplet style encoder instead of triplet only (ablation) | DiffusionPen · IAM writer classification | accuracy (%): 67.50 -> 70.31 | Improved |
| use hybrid style encoder instead of classification only for FID (ablation) | DiffusionPen · IAM generated words | FID: 22.23 -> 22.54 | No clear effect |
| use hybrid style encoder instead of classification only for HTR imitation (ablation) | DiffusionPen · HTR on synthetic IAM | CER (%): 7.12 -> 6.94 | Improved |
| condition on few-shot style embedding from 5 samples instead of class embedding (WordStylist) | latent diffusion HTG · HTR on synthetic IAM | CER (%): 8.26 -> 6.94 | Improved |
| add small style-embedding noise (0.25) (ablation) | DiffusionPen · HTR on synthetic IAM | CER (%): 6.94 -> 6.79 | Improved |
| increase style-embedding noise to 2.00 (ablation) | DiffusionPen · HTR on synthetic IAM | CER (%): 6.79 -> 7.56 (0.25 vs 2.00) | Hurt |
| add noise bias in the diffusion prior (ablation) | DiffusionPen · HTR on synthetic IAM | CER (%): 6.94 -> 6.93 | No clear effect |
| augment real IAM training with 55K DiffusionPen samples | CNN-LSTM HTR · IAM word-level HTR | CER (%): 5.14 -> 4.71 | Improved |
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.
- GANwriting — GANwriting: Content-Conditioned Generation of Styled Handwritten Word Images (ECCV 2020) · better on 3 of 3 shared results
- VATr — VATr++: Choose Your Words Wisely for Handwritten Text Generation (TPAMI 2024) · better on 1 of 1 shared result
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
- Beyond Isolated Words: Diffusion Brush for Handwritten Text-Line Generation (ICCV 2025 · similarity 0.92)
- Zero-Shot Styled Text Image Generation, but Make It Autoregressive (CVPR 2025 · similarity 0.87)
- Decoupling Layout from Glyph in Online Chinese Handwriting Generation (ICLR 2025 · similarity 0.86)
- Layout-Agnostic Scene Text Image Synthesis with Diffusion Models (CVPR 2024 · similarity 0.85)
- MetaWriter: Personalized Handwritten Text Recognition Using Meta-Learned Prompt Tuning (CVPR 2025 · similarity 0.85)
- FontDiffuser: One-Shot Font Generation via Denoising Diffusion with Multi-Scale Content Aggregation and Style Contrastive Learning (AAAI 2024 · similarity 0.85)
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.
