Open Mr-Nobody-dey opened 11 months ago
why not show your results? so that the author can figure it out,
I deleted all the results. I will try running the script again and upload the results when i will have some time.
I think similar issue is also raised in issue no "https://github.com/wl-zhao/DiffSwap/issues/6#issue-1909889749". I also had similar results. Basically in the output of swapped face, the identity of the source face is not clearly visible.
I tested the model, and the swap quality is very poor. What might be the possible reason? Do I need to train it on source and target datasets? Or it works only on trained datasets?