speech-driven-animation icon indicating copy to clipboard operation
speech-driven-animation copied to clipboard

How long does it take to run?

Open jdchanggg opened this issue 6 years ago • 4 comments

I thought it shouldn't take long, but this line:

va = sda.VideoAnimator(gpu=0) # Instantiate the animator

has been running for over an hour. Is that supposed to be right?

jdchanggg avatar Dec 10 '19 19:12 jdchanggg

Instantiating the animator shouldn't take long since all it does is load the network weights. Maybe something is wrong in your GPU. Have you tried loading it on cpu (with gpu=-1) to compare? The whole process including the generation of the video shouldn't take more than a few seconds.

DinoMan avatar Dec 12 '19 15:12 DinoMan

@jdchanggg did you solve the problem? i meet the same question, it keeps showing "Downloading the face detection CNN. Please wait..." and result nothing else. There is noting wrong with my GPU

ustc-baize avatar Dec 15 '19 14:12 ustc-baize

@jdchanggg did you solve the problem? i meet the same question, it keeps showing "Downloading the face detection CNN. Please wait..." and result nothing else. There is nothing wrong with my GPU

Ok I see what is wrong. My library uses the face-alignment library package from: https://github.com/1adrianb/face-alignment The face alignment library, when used for the first time, will download the model and apparently it is failing to do so. I am not the maintainer of the face-alignment library so I can't actually fix the issue.

However, you can align your faces through another means and use the align=False option when running the video animator. The mean face I use is loaded in the code so you can use that to align your faces.

DinoMan avatar Dec 16 '19 13:12 DinoMan

install the facealignmet using pip ? or download it from https://pypi.org/project/face-alignment/

ShimYang avatar Dec 19 '19 09:12 ShimYang