Closed yashfk001 closed 6 years ago
In the first instance, I'd check your cuda install is working correctly:
https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#verify-installation
Cuda and cudnn installation is proper. I am able to train older extracted pictures with gpu. Train seems to work properly, issue is there in extract and convert. It is not detecting any faces
Seems that a filter image is specified. What is in this image ?
I cant find this image and i didn't apply any filter on it, i just ran "python faceswap.py extract -i photo/cage -o photo/cageOut". How can i remove this filter?
The filter shouldn't matter. It's just the default value.
Honestly, when I had this issue, it was a dlib/cuda issue. Nothing to do with faceswap per se. Just because it trains does not mean your install is ok, as DLIB and Tensorflow have different implementations.
@torzdf : Issue was with dlib, now that is resolved. Its working. Thanks @torzdf , @Clorr
Extract in not detecting any faces in any of the images, i have used your reference Nicolas cage pictures for this. It wont detect face on any of them. I have gone through all issues submitted here and tried to resize images or try with new ones, but it didnt work