This project aims to automatically animate SignWriting into skeletal poses.
This is the reverse of signwriting-transcription.
We would ideally like to have to implementations:
- diffusion - a diffusion-based method to animate SignWriting
- translation - a translation-based method to animate SignWriting using the sign-vq model.
pip install git+https://github.com/sign-language-processing/signwriting-animationTo animate a SignWriting FSW sequence into a .pose file:
signwriting_to_pose --signwriting="M525x535S2e748483x510S10011501x466S2e704510x500S10019476x475" --pose="example.pose"When generating full sentences, it is recommended to post-process the .pose files using
fluent-pose-synthesis.
The full data is available in sign/data. These examples are taken from the DSGS Vokabeltrainer:
| 00004 | 00007 | 00015 | |
|---|---|---|---|
| SignWriting |  |
 |
 |
| Video |  |
 |
 |