Research map / Image generation / Text-to-image diffusion
GANwriting: Content-Conditioned Generation of Styled Handwritten Word Images
Lei Kang, Pau Riba, Yaxing Wang, Marçal Rusiñol, Alicia Fornés, Mauricio Villegas
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 writer classification loss to the discriminator loss (ablation) | GANwriting (Ld only) · Learning objectives, content 'vision' | FID: 364.10 -> 207.47 | Improved |
| add recognition loss to the discriminator loss (ablation) | GANwriting (Ld only) · Learning objectives, content 'vision' | FID: 364.10 -> 138.80 | Improved |
| add recognition loss to Ld + Lw (ablation) | GANwriting (Ld + Lw) · Learning objectives, content 'vision' | FID: 207.47 -> 130.68 | Improved |
| add writer classification loss to Ld + Lr (ablation) | GANwriting (Ld + Lr) · Learning objectives, content 'vision' | FID: 138.80 -> 130.68 | Improved |
| use out-of-vocabulary words instead of in-vocabulary words (seen styles) | GANwriting · IAM per-writer FID | FID: 120.07 -> 125.87 | Hurt |
| use unseen styles instead of seen styles (in-vocabulary words) | GANwriting · IAM per-writer FID | FID: 120.07 -> 124.30 | Hurt |
| use both OOV words and unseen styles | GANwriting · IAM per-writer FID | FID: 120.07 -> 130.68 | Hurt |
| generated vs genuine images in human study | GANwriting · Human evaluation | examiner accuracy: 49.3 | No clear effect |
| train all networks jointly from scratch instead of pre-training W and D (ablation) | GANwriting · training strategy | image quality | Improved |
Later papers that compared with this method
3 later accepted papers report GANwriting as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- DiffusionPen: Towards Controlling the Style of Handwritten Text Generation (ECCV 2024 · better on 3 of 3 shared results)
- Handwriting Transformers (ICCV 2021 · better on 5 of 5 shared results)
- HiGAN: Handwriting Imitation Conditioned on Arbitrary-Length Texts and Disentangled Styles (AAAI 2021 · 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
- Handwritten Text Generation From Visual Archetypes (CVPR 2023 · similarity 0.90)
- HiGAN: Handwriting Imitation Conditioned on Arbitrary-Length Texts and Disentangled Styles (AAAI 2021 · similarity 0.89)
- VATr++: Choose Your Words Wisely for Handwritten Text Generation (TPAMI 2024 · similarity 0.88)
- Zero-Shot Styled Text Image Generation, but Make It Autoregressive (CVPR 2025 · similarity 0.88)
- Learning to Generate Stylized Handwritten Text via a Unified Representation of Style, Content, and Noise (ICLR 2026 · similarity 0.88)
- Rich Human Feedback for Text-to-Image Generation (CVPR 2024 · 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.
