Closed hacker009-sudo closed 7 months ago
Please fill in a proper issue - specifically:
In general, the stylegan3
code is still work-in-progress, but if I understand what you are trying to do I might be able to help.
@AmitMY , I am following the documentation to run the code for pose_to_video and running the below command on colab:
!python /content/transcription/pose_to_video/bin.py --model=stylegan3 --pose=sign.pose --video=sign.mp4
Issue:
Loading input pose ...
Generating video ...
Traceback (most recent call last):
File "/content/transcription/pose_to_video/bin.py", line 58, in
I have even tried with pix_to_pix but same issue:
ModuleNotFoundError: No module named 'pose_to_video.pix_to_pix'
Request you to please help
Thank you for your explanation. It is now fixed, and the commands do run, for example:
pose_to_video --model=pix_to_pix --pose=sign.pose --video=sign.mp4 --upscale
Please understand however, that what you are running is under "Next steps" which is NOT ready. For example, this pix_to_pix model yields very bad results, and I plan to rework it in the next month or so, as well as the stylegan model.
EDIT: Sorry, this isn't working right now actually because of too-large model files. Instead of fixing those right now with LFS, since the output quality is anyway very very bad, I will fix them later differently.
Thanks for the explanation. Will wait for the final implementation.
I had some time yesterday to re-train and fix some issues https://github.com/sign-language-processing/transcription/commit/a308b5868f16290623f3f0f408b6a151905e0e82
The output is now more reasonable. I'll let it train for another day or two, and then deploy it
Pose to video moved to https://github.com/sign-language-processing/pose-to-video
@AmitMY , please help me to resolve this issue