Dataset
bcba25 opened this issue · comments
Could you reupload dataset anywhere else, baidu is too slow. Thanks
I've uploaded it to google drive.
https://drive.google.com/file/d/1CUG9HK75oGYxQn81RU8H-zjscZy8ozMp/view?usp=sharing
@liurs1990 alright thanks! Btw which camera is used to capture depth data ?
@hxdef2517 It's RGB face anti-spoofing without depth camera. You can get the depth information of RGB frame by PRNET.
@liurs1990 I see, got it. I've read the paper but can't get one simple thing ) when we do training and set depth map to 0 for all fake images, what does this gives us when we try to predict on trained model ? Is it something like, real images got a good depth but then when we try to predict on fake on it will not have that good depth comparing to the real one? Could you explain this particular part please )))
Yes, the ideal result is getting a depth image on the real face and dark image on the fake face, but the depth image isn't the final result. We need the score of classifier instead of depth image.
@liurs1990 alright, thanks for clarification and sharing the codes