Page 1 of 1
Converting A->A or B->B
Posted: Mon Jul 26, 2021 8:20 am
by apueyo
Hi, is it currently possible to run a conversion process in which to produce face A from A, or face B from B, on a set of faces or image sequence, with an already trained model?
That would potentially allow for good comparisons and maybe an extra step of refinement via an external neural network.
Thank you!!
Re: Converting A->A or B->B
Posted: Mon Jul 26, 2021 8:51 am
by torzdf
Sure. Select the swap
option in convert settings to swap which decoder is used.
Re: Converting A->A or B->B
Posted: Mon Jul 26, 2021 9:07 am
by apueyo
Hey, wouldn't that convert B into A?
What I'd be looking for is encoding A and decoding it again into A. Or encoding/decoding B into B.
Thanks!!
Re: Converting A->A or B->B
Posted: Tue Jul 27, 2021 9:29 am
by torzdf
Well.... It would only change if from B to A if you started sending it B's faces.
By swapping the model you our swapping the decoder, so you will get "A"s decoder rather than "B"'s
Re: Converting A->A or B->B
Posted: Wed Jul 28, 2021 5:12 pm
by apueyo
Thank you, it does work!!!!
- Edit: I finally understood the encoder is shared and it's only the decoder that is swapped. Sorry for asking about something already documented - hopefully this question is useful for somebody else still grasping these concepts