search

mapooon / BlendFace

[ICCV 2023] BlendFace: Re-designing Identity Encoders for Face-Swapping https://arxiv.org/abs/2307.10854
Other
164 stars 7 forks source link

Results of this model. #11

Open Mr-Nobody-dey opened 5 months ago

Mr-Nobody-dey commented 5 months ago

The swap result of this model is not good. The similarity of the source face with the swapped face is very poor.

Note that-> I have used proper alignment methods as mentioned in this repository.

I will upload the similarity matrix and result later.

linzixuan45 commented 5 months ago

I also encountered the same problem. I tried to use BlendFace weights to train the face swap model, but the results showed that the identity similarity between Source and Swap was low. The swap matrix results during training were also not very good (as shown below. The left side is the Target, and the upper side is the Source). In addition, I set the training step to 50,000 steps, batch size to 16, and learning rate to 0.0001.

linzixuan45 commented 5 months ago

![Uploading step_49200.jpg…]()

linzixuan45 commented 5 months ago

step_49200

YoucanBaby commented 2 months ago

same question