sign-language-processing/datasets

PSL Dataset

imsamimalik opened this issue · 2 comments

I visited https://sign.mt/ and I saw Pakistan Sign Language listed there as one of the options for signed languages. But here I do not see any dataset for PSL. How are you training the model for PSL?

Is the 3D avatar and human model broken or it is not working on my side?

AmitMY commented

For PSL, we are leveraging https://spreadthesign.com/ur.pk/search/? under a special research license.
There is no training of any model involved, only https://github.com/ZurichNLP/spoken-to-signed-translation sign stitching.
Since the license does not allow, this dataset is not included here.

(As for your question about the https://github.com/sign/translate project, the 3d Avatar is not functional, but the human model should be. If you have any specific issue, please file it in that project)