This is the repository for the paper: SignSynth: Data-Driven Sign Language Video Generation.
gloss2pose was originally written in tensorflow and later ported to pytorch. For the full 2D data, please contact me under s.m.stoll@surrey.ac.uk. I cannot grant you access to the actual SMILE dataset as I am no longer associated with the owners of the dataset. Please contact n.camgoz@surrey.ac.uk for enquiries regarding this dataset.
I will clean up this repository in the next few weeks.
If you are using the code please cite:
@inproceedings{Stoll_ACVR2020,
author = {Stoll, Stephanie and Hadfield, Simon and Bowden, Richard},
title = {SignSynth: Data-Driven Sign Language Video Generation},
booktitle = {ECCV 2020 Workshop Proceedings},
year = {2020}
}
pose2video includes files from pix2pixHD. Please check out their repository: https://github.com/NVIDIA/pix2pixHD