👹 Liveness detection (aka AttackNet)

🗒️ Introduction

Code and materials for a research paper about liveness detection, originated from the Andrea Maranesi's research.

The idea of this project is to create a model, which we call AttackNet, that can detect if the given face is fake or not.

🏗️ Model architecture

🗃️ Datasets used

For this research paper we used five different datasets:

• CSMAD
• [3DMAD]
• [MSSPoof]
• [Replay-Attack]
• Our Dataset

Some of the example images from the datasets are shown below.

☑️ Cross-Validation Results

To test our model, we trained the neural network on one dataset and then assessed it on the other one. Results are shown below.