Patch-Based Spoof Detection

Patch-Based Spoof is a deep learning framework for face anti-spoofing using patch-based convolutional neural networks. This repository provides modular code for training, evaluating, and experimenting with patch-based and keypoint-based models for presentation attack detection (PAD).

Features

• Patch-based CNN models for spoof detection
• DeepPix and its variants (HSV, six-channel, extended)
• Keypoint-based experiments
• Modular dataloaders and trainers
• Evaluation scripts and metrics (APCER, BPCER, ACER)
• Jupyter notebooks for exploration and visualization

Installation

1. Clone the repository:

   git clone https://github.com/mdhasanai/patch-based-spoof.git
   cd patch-based-spoof

2. (Recommended) Create a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

   (Create a requirements.txt if not present, listing packages like torch, torchvision, numpy, pandas, etc.)

Usage

Training a Patch-Based Model

python scripts/patch_train.py --config <config-file>

Training DeepPix Model

python scripts/deeppix_train.py --config <config-file>

Evaluation

python scripts/eval_keypoint.py --model <model-path> --data <data-path>

Visualization

Explore the provided Jupyter notebooks in scripts/ for filter analysis, heatmap visualization, and more.

Project Structure

patch-based-spoof/
├── dataloaders/         # Data loading utilities
├── models/              # Model architectures (DeepPix, Patch-based CNN)
├── scripts/             # Training, evaluation, and analysis scripts
│   ├── metrics/         # Metric calculation scripts
│   ├── eval/            # Evaluation utilities
├── trainers/            # Training logic for different models
├── utils/               # Helper functions (evaluator, etc.)
├── README.md            # Project documentation

Training & Evaluation

• Configure your training and evaluation parameters in the respective scripts or config files.
• Use the provided scripts for patch-based or keypoint-based experiments.
• Metrics such as APCER, BPCER, and ACER are available in scripts/metrics/.

Contributing

Contributions are welcome! Please open issues or pull requests for bug fixes, new features, or improvements.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

• Inspired by DeepPix and related face anti-spoofing research.
• Thanks to all contributors and the open-source community.

---

For questions or support, please open an issue on GitHub.