search

liangchen527 / SLADD

Official code for Self-supervised Learning of Adversarial Example: Towards Good Generalizations for Deepfake Detection (CVPR 2022 oral)
Apache License 2.0
131 stars 16 forks source link

Generation of composite image #13

Open zpshs opened 1 year ago

zpshs commented 1 year ago

How to generate composite pictures

ucalyptus2 commented 1 year ago

@zpshs what do u mean by that? please explain so that i can understand if it is a useful thing.

zpshs commented 1 year ago

I want to output synthesized pictures during training

zpshs commented 1 year ago

And I want to know, what is the specific role of mask? thank you

zpshs commented 1 year ago

If I change the dataset, how can the mask get it? Is mask the face position of the faked image?

zpshs commented 1 year ago

I mean, I want to take out the face of the image first and train it. But I don't know how to handle the mask after doing this.