search

OeslleLucena / FASNet

CNN archictecture for face anti-spoofing
MIT License
171 stars 56 forks source link

I wonder why not process the depth data in image? #4

Open SaSaNamiRinSage opened 6 years ago

SaSaNamiRinSage commented 6 years ago

The project achieved 100% accuracy of 3DMAD, and only used the color image for training. Wonder how much the depth data will help for further improvement. For example, why iPhone X's face recognition uses depth image.

Thanks

OeslleLucena commented 6 years ago

Hi @SaSaNamiRinSage , sorry for the late response. The depth will give information among the frames, so it is expected that improve the accuracy of the REPLAY-ATTACK or other datasets for example. the 3D MASK dataset was an easy one for the CNN.