Deepfake Detection Application 🕵️‍♀️🤖

Project Overview

This project implements deepfake detection system using a combination of Xception neural network architecture and LSTM (Long Short-Term Memory) networks. The application provides a user-friendly Streamlit interface for real-time deepfake image detection.

🌟 Key Features

• Advanced Deep Learning Model: Utilizes Xception as the base feature extractor
• Temporal Analysis: Incorporates LSTM for enhanced temporal feature recognition
• High Accuracy Deepfake Detection: Robust model trained on diverse deepfake datasets

🛠 Technology Stack

• Deep Learning: TensorFlow, Keras
• Web Interface: Streamlit
• Data Processing: NumPy, Pandas
• Image Processing: OpenCV
• Visualization: Matplotlib, Seaborn

Dataset

I used 600 videos each class (real and fake) of CelebDF V2, you can access it here https://github.com/yuezunli/celeb-deepfakeforensics

🚀 Installation

1. Clone the repository:

   git clone https://github.com/farhanharvito/DeepfakeDetection
   cd deepfake-detection

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

🖥 Running the Application

To launch the Streamlit application:

streamlit run app.py

You can download the pre-trained model here

📊 Model Performance

Metrics

• Accuracy: 91.67%
• Precision: 89.06%
• Recall: 95%
• F1 Score: 91.64%

🔍 How It Works

1. Feature Extraction:

   • Xception network extracts deep visual features from input images
   • Handles complex spatial patterns in potential deepfake images

2. Temporal Analysis:

   • LSTM layers process extracted features
   • Captures temporal dependencies and sequence-level information

3. Classification:

   • Final dense layers make binary classification (Deepfake vs. Real)

📋 Requirements

See requirements.txt for the complete list of dependencies.

🧾 License

Distributed under the MIT License. See LICENSE for more information.

🔬 Research and References

Research Problem References

• Quick Classification of Xception And Resnet-50 Models on Deepfake Video Using Local Binary Pattern
• A Comparative Study of Deepfake Video Detection Method
• Celeb-DF: A Large-scale Challenging Dataset for DeepFake Forensics

Learning References

• Xception: Deep Learning with Depthwise Separable Convolutions
• Long Short-term Memory RNN

Disclaimer: This project is for educational and research purposes. Always use AI responsibly.