You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello this is great work!!how do you feed Multiple identities into the attention module when obtaining multiple id tokens through id extractor in multiple identities applications? concat directly behind prompt embeds?
The text was updated successfully, but these errors were encountered:
@Daming-TF Here is my guess: During training, they used raw images with the face region masked out as the stylized human image, while the cropped face from the same image acts as the identity input.
We can follow the same method for inference: Suppose we want to put the faces of individuals A and B on C and D within a reference image respectively. First, we use an instance segmentation model to create a mask for C in the reference image. Then, we used the facial features of individual A as a guiding reference.
Hello this is great work!!how do you feed Multiple identities into the attention module when obtaining multiple id tokens through id extractor in multiple identities applications? concat directly behind prompt embeds?
The text was updated successfully, but these errors were encountered: