search

ee09115 / spoofing_detection

Cork/Face Presentation Attack Detection
342 stars 98 forks source link

Training #2

Closed rebotnix closed 5 years ago

rebotnix commented 5 years ago

Hi,

i would like to produce some own datasets with webcam and own print magazins.

Do you can show us how you train the models by providing the training script?

Thx.

ee09115 commented 5 years ago

Hello,

I have added the script I used to train the ETC classifier.

The input file is a csv file which have the following format:

<feature vector 1>,<label 1>,<PAI> <feature vector 2>,<label 2>,<PAI> ... <feature vector N>,<label N>,<PAI>

The feature vector is a floating point vector, FV=(Y,Cr,Cb,Luv), which contain the frequency histogram normalized in the range [0.0, 255.0]. The label is the corresponding label to the nature of the feature vector: I used 1 for an attack, and 0 for a bona fide image. The PAI, is the Presentation Attack Instrument used to produce the attack. It is useful when you want to use ACER metric to evaluate your PAD instrument.

Regards,

rebotnix commented 5 years ago

Great!! Thank you very much.

DecentMakeover commented 5 years ago

Hi

Thanks for sharing your work, but how did you compute the feature vector for an image?

Thanks in advance