This is a permissions error on your model folder:
OSError: Unable to create file (unable to open file: name = 'D:\_own_folder_\FaceSwap\Donald Trump - Elon Musk\Model\original_decoder_A.h5', errno = 13, error message = 'Permission denied', flags = 13, o_flags = 302)
The .fsa file is generated at the extract stage and will be placed with the video that you extracted from.
This, unfortunately, sounds like a hardware/power issue.
In the first instance, make sure all GPU overclocks are disabled.
Beyond that, it will be next to impossible to troubleshoot hardware issues remotely.
You're going to have to give a bit more information than that. What are you trying to do? Import where? What command are you running? What are you seeing?
Your best bet is to search for Autoencoders and read up on that.
Then try to apply that with what we do (i.e share the encoder, then swap the decoders on convert)
The model doesn't contain alignments. It just contains algorithms for decoding and encoding face images.
You don't run conversion on extracted faces. You run conversion on videos/full frame images
I believe the flipping is in the Umeyama code. You are welcome to have a look to diagnose a problem/find a fix:
Umeyama:
https://github.com/deepfakes/faceswap/b ... umeyama.py
Alignment:
https://github.com/deepfakes/faceswap/b ... aligner.py
I have never seen this before, and this looks fine on the face of it....
Triple check your paths, and check permissions on the model folder?
For the models, you can download the zip files and extract them to the following locations:
Aligners - <faceswap_folder>/plugins/extract/align/.cache
Detectors - <faceswap_folder>/plugins/extract/detect/.cache
Maskers - <faceswap_folder>/plugins/extract/mask/.cache
Generally the extended mask fixes it in most issues. Use the preview tool to adjust the mask erosion and blending settings to get it how you need it.
Unfortunately I'm not the person to ask. Hopefully [mention]bryanlyon[/mention], or someone who uses it will be along to offer some tips.
I'm not sure what the issue is here. Can you be a bit more specific? Ultimately the model needs to be trained more.
I'm personally of the opinion that pre-training offers little to no benefit.
The basic concept is to start training a model on lots of random faces so it learns what a face is, prior to moving over to your actual training set.
The video that you are trying to convert for needs an alignments file generated so that it knows where the faces are.
For some reason you are running the wrong version of Keras.
Currently, the highest supported version of Keras is 2.2.4