Amit Moryossef

By optimizing in my case I meant early-stopping. I'll try to add to the decode and evaluate and see if I can then early stop on those

Thanks. I think that since my implementation is generic for SignWriting outputs, it might be valid to introduce inside of sockeye. It does not change the behavior for spoken language...

Signed-to-spoken translation is not yet implemented in the app. The graph in the wiki explains what is implemented and what is not: https://github.com/sign/translate/wiki/Signed-to-Spoken Yellow means: implemented but does not work...

(Related, but will not solve, https://github.com/firebase/firebase-tools/issues/6742)

Repository to get this done https://github.com/sign-language-processing/signwriting-illustration

For the current system - do you have a dictionary with the signs in the language?

Seems like there are 1092 videos, covering a smaller vocabulary (multiple videos per word). Also seems like there is an "example sentence", but one must login to see it. If...

Unfortunately, their font is not standard, and I only use SignWriting. So we will need permission from them to use their videos, if they give us permission I could load...

They don't seem to have SignWriting, but yes, we will need permission. If you want to serve the app with SSL locally, follow https://medium.com/@mariokandut/how-to-configure-ssl-locally-in-angular-413077568caa Of course, on https://sign.mt/ it is...

For other issues, please open another issue, and include a screen-recording video showing what you see