search

neverUseThisName / Face-X-Ray

Unofficial implementation of paper 'Face X-ray for More General Face Forgery Detection'. (updating...)
91 stars 14 forks source link

Would you mind providing the database? #2

Open zhuhongwei1999 opened 4 years ago

zhuhongwei1999 commented 4 years ago

Hi, I run your code under LFW-a database of face photographs, but the result is not as good as yours. Would you mind providing the link of your face database? Thank you.

neverUseThisName commented 4 years ago

I don't think it has anything serious to do with dataset, but I'm not sure. Would you mind providing some samples of bad results? There are also some parameters you can tune:

  1. The threshold for landmark distance.
  2. Change how you calculate landmark distance.
  3. The random deform std (Line 146).
  4. Gaussian blur kernel size and std (Line 71)
zhuhongwei1999 commented 4 years ago

forge_1 forge_11

Most of my results are weird because of improper color transformation of target image.

neverUseThisName commented 4 years ago

The problem might be that color_transfer estimates the color style of the whole image. If there are distinct components (e.g. foreground and background) as is your case, the resultant style would be the neutralization of them. Maybe you can apply color_transfer only on face regions (e.g. cropping the face region).

bobleono commented 3 years ago

hello,why do i get the result that no math is in the datasheet and i did not get any result in the result file