Anti-Spoofing Detector

The AntiSpoofingDetector is an advanced anti-spoofing/liveness detection system that leverages facial recognition technology. This project is implemented using Python and OpenCV, providing a robust solution to distinguish between real faces and spoof attempts.

Features

• Facial Recognition: The detector utilizes state-of-the-art facial recognition techniques to identify and authenticate real faces.

• Liveness Detection: With the help of a trained model, the system assesses the liveness of detected faces to prevent spoofing attempts.

• Python and OpenCV: The project is implemented in Python, a versatile and widely-used programming language, and employs the OpenCV library for computer vision tasks.

• Trained Model: The AntiSpoofingDetector incorporates a pre-trained model, ensuring accurate and reliable performance in different scenarios.

Installation

To use the AntiSpoofingDetector, you need to install the following dependencies:

OpenCV

pip install opencv-python

Ultralytics

pip install ultralytics

Mediapipe

pip install mediapipe

PyTorch (Optional, for GPU acceleration)

If you have a compatible GPU and want to accelerate the system using CUDA, follow these steps:

1. Install the CUDA Toolkit: CUDA Toolkit Installation Guide.

2. Install PyTorch with GPU support: https://pytorch.org/get-started/locally/

Dataset Setup

To set up the Dataset directory and its subdirectories for the AntiSpoofingDetector, follow these steps:

1. Create the Dataset Directory:

Open a terminal or command prompt and navigate to the root directory of the AntiSpoofingDetector project. Run the following command:

mkdir Dataset

2. Create Subdirectories:

Open a terminal or command prompt and navigate to the root directory of the AntiSpoofingDetector project. Run the following command:

mkdir Dataset/all
mkdir Dataset/DataCollect
mkdir Dataset/Fake
mkdir Dataset/Real
mkdir Dataset/SplitData

3. Populate the Directories

If you choose to manually populate the dataset, consider adding real faces and fake faces to the Dataset/all directory, aiming for an equal split of each.

Before proceeding, make sure to alter the classID parameter in the dataCollection.py file:

• Set classID to 0 for fake faces.
• Set classID to 1 for real faces.

This adjustment ensures proper labeling during the data collection process.

Note: It's essential to maintain a balanced representation of real and fake faces in the Dataset/all directory to enhance the performance and reliability of the AntiSpoofingDetector.

After making these changes, you can proceed with the data collection process as outlined in the project instructions.

Models

The AntiSpoofingDetector relies on various archives for its functionality. Follow these steps to set up the necessary components:

1. Download the Models

Obtain the models archive file named models.zip from the provided source.

2. Extract the Models

Locate the downloaded models.zip file in your file system.

On Windows:

• Right-click on the models.zip file.
• Select "Extract All..." and choose the destination folder.

On Linux or macOS:

• Open a terminal.
• Navigate to the directory containing the models.zip file.
• Run the following command:

  unzip models.zip

By following these steps, you'll have successfully extracted the necessary components for the AntiSpoofingDetector, including the dataset, runs, and models. Adjust the instructions as needed for your specific project structure and user requirements.

Running the Application

To run the AntiSpoofingDetector, follow these steps:

1. Open a terminal.
2. Navigate to the directory containing main.py:

   cd /path/to/AntiSpoofingDetector

3. Run the main script:

   python main.py

4. To quit press:

   q

Training a New Model

To train a new model, follow these steps:

1. Navigate to the directory containing train.py:

   cd /path/to/AntiSpoofingDetector

2. Run the training script:

   python train.py

3. After training, use the best model (best.pt) from run/detect/train/weights and place it in the models folder.
4. Use the trained model in main.py for liveness detection.