rodrigobressan / padify

Framework to perform PAD (Presentation Attack Detection) on Facial Recognition systems through intrinsic properties and Deep Neural Networks - Still Under Development

PADify (pronounced as /ˈpædfaɪ/) is a project to perform presentation attack detection (PAD) on publicly available datasets by using different intrinsic image properties along with Convolutional Neural Networks.

Presentation Attack Detection

_{Fig 1 - Examples of Presentation Attack Detection}

Pipeline Overview

_{Fig 2 - Overview of proposed method}

Properties

• Illuminant Maps (Carvalho et al.)
• Saliency (Zhu et al.)
• Depth (Godard et al.)

Datasets

• CASIA Face Anti-Spoofing Database
• Replay Attack
• NUAA Imposter Database
• ROSE-Youtu Face Liveness Detection Dataset (In Progress)

License

This file is licensed under the MIT license. You can check the LICENSE file for more information.

padify's Issues

Is it possible to share the pretrained network?

Hi @rodrigobressan
I am interested in the project and I wonder if you have any good pretrained network for test and evaluating?

Many thanks

How does it run?

I'm interested in your project. How does it run?