dailenson/SDT
A Transformer-based model that disentangles writer-wise and character-wise styles to generate realistic handwriting.
Not currently ranked — collecting fresh signals.
star history
The Style-Disentangled Transformer (SDT) separates writer and character style representations to generate online handwriting with conditional content and style. The model uses contrastive learning and Gaussian mixture models to disentangle subtle style variations between characters written by the same person. It extends to offline Chinese handwriting generation and was published at CVPR 2023.
Frequently asked
- What is dailenson/SDT?
- A Transformer-based model that disentangles writer-wise and character-wise styles to generate realistic handwriting.
- Is SDT open source?
- Yes — dailenson/SDT is open source, released under the MIT license.
- What language is SDT written in?
- dailenson/SDT is primarily written in Python.
- How popular is SDT?
- dailenson/SDT has 1.4k stars on GitHub.
- Where can I find SDT?
- dailenson/SDT is on GitHub at https://github.com/dailenson/SDT.