Deepfake Detection System
Welcome to the Deepfake Detection System, a cutting-edge project designed to identify video deepfakes using advanced deep learning techniques. By combining ResNext for feature extraction and LSTM for sequence analysis, this project achieves high-accuracy detection of manipulated videos.
🚀 Features
Robust Deepfake Detection: Leverages a pretrained ResNext CNN and LSTM for accurate classification. User-Friendly Web App: Built with Django, allowing users to upload videos and view results seamlessly. Dockerized Deployment: Spin up the application effortlessly with Docker, no dependency hassles. Detailed Documentation: Comprehensive guides to understand and replicate the project.
📂 Project Structure Deepfake-Detection-System ├── Django Application # Web interface for video uploads and predictions ├── Model Creation # Scripts for building and training the model ├── Documentation # In-depth project guides and resources
🛠️ System Architecture The system extracts features from video frames using a pretrained ResNext model, which are then processed by an LSTM network to classify videos as real or fake.
⚙️ Installation
-
Clone the Repository git clone https://github.com/viraj7066/Deepfake-Detection-System.git cd Deepfake-Detection-System
-
Using Docker
Ensure Docker is installed. Build and run the container:docker build -t deepfake-detection . docker run -p 8000:8000 deepfake-detection
Access the app at http://localhost:8000.
- Manual Setup
Follow the YouTube Installation Playlist for step-by-step setup instructions.
🖥️ Usage
Navigate to the Django Application directory. Start the Django server:python manage.py runserver
Open your browser and visit http://localhost:8000. Upload a video to detect if it’s a deepfake.
📚 Documentation
Project Documentation Medium Article for an in-depth overview
🤝 Contribute We welcome contributions to improve the project! Suggested enhancements include:
Deploying to free cloud platforms. Creating an open-source API for detection. Supporting batch processing of entire videos. Optimizing code Wfor faster execution.
Completed Improvements:
✅ Dockerized the application. ✅ Enabled compatibility with non-CUDA systems (e.g., AMD GPUs, CPUs).
To contribute:
Fork the repository. Create a feature branch. Submit a pull request.
📜 License This project is licensed under the GPLv3 License.
⭐ Support If you find this project valuable, please give it a star on GitHub to show your support!