Faceswap Forum

Ah perfect.
Missed it.
Thnx!

Is it possible to do a reverse swap?
So I train face B on top of face A now, but can I switch that and use the same model to put face a on face b?
In essence all that is needed is in the model..

The settings are a bit confusing tbh.
If I set Shared Fc to none but leave Split Fc unchecked, do I have a Shared Fc or not?

Is it possible to only have Gblock as an Fc and turn the regular Fc off?
If Shared Fc is set to None a Split Fc is created by default I noticed.

Does it makes sense to have a higher output size then the input layer of the encoder?
Like when I set output size to 1024 but scale the fs_original encoder to 25 (512), is the output actually gonna be 1024?
Or is it in essence training on 512 and upscaling it to 1024 with no extra detail.

I just started running dny 1024 on 48gb with Full fc, Split fc, Split Gblock and Split Decoders at bs 12.
Little stunned myself tbh.

Interesting.

Besides output size, does the max input resolution of the encoder also have an effect on the resizing?
I see efficientnet_v2 had a max input of 480px.
So I assume the input to the encoder is resized to 480px if I feed it higher resolution faces.
Even if I have an output size of 512.

I ask because this is the case in DFL.

Can someone explain why there is the option to adjust the convert batchsize in the global training settings?
Has this to do with the convert that is showing in the preview?

I see. Thanks for clearing that up.

Then, just to see if I understand the NN correctly;
Is the first layer of nodes in the encoder the amount of pixels of the image times 3? (for each rgb channel)
So that each node contains an r, g or b value of a single pixel?

Yes, I checked the state.json file.
But they are not the same as in the Training settings in the GUI.
I guess I should have saved them as a preset upon creation.