Deep Pixel-wise Binary Supervision for Face PAD

This package is part of the signal-processing and machine learning toolbox Bob.

This package contains source code to replicate the experimental results published in the following paper:

@inproceedings{GeorgeICB2019,
    author = {Anjith George, Sebastien Marcel},
    title = {Deep Pixel-wise Binary Supervision for Face Presentation Attack Detection},
    year = {2019},
    booktitle = {ICB 2019},
}

If you use this package and/or its results, please consider citing the paper.

The main idea in the paper is the use of pixel-wise binary supervision in addition to binary supervision. The pixel-wise labels are illustrated below.

Fig. 1 Figure showing pixe-lwise binary labels.

Users Guide

• Running DeepPixBiS on OULU and REPLAY-MOBILE Datasets
  • Running experiments with DeepPixBiS architecture in OULU-NPU Protocol_1
  • Running experiments with DeepPixBiS architecture in REPLAY_MOBILE grandtest protocol
  • Running cross DB experiments with DeepPixBiS
  • Using pretrained models
• Running baseline experiments in OULU_NPU database
  • Running experiments in OULU_NPU database
  • Running experiments in REPLAY-MOBILE database
• Running baseline experiments in Cross database settings
  • Training on REPLAY-MOBILE and testing on OULU-Protocol_1
  • Training on OULU-Protocol_1 and and testing on REPLAY-MOBILE
• References