DeepTrace

Modular High-Accuracy Deepfake Generation Pipeline for Academic Research

Installation

This is a Final Year Project (FYP) focused on deepfake generation for academic research. The installation requires Python 3.10+ and technical knowledge.

Quick Start

1. Clone the repository:

   git clone https://github.com/Ali7040/Deepfake-gen-pipeline.git
   cd Deepfake-gen-pipeline

2. Install dependencies:

   python install.py

3. Run the application:

   python deeptrace.py run

For detailed setup instructions, see QUICK_START.md.

Usage

Basic Commands

# Run with GUI
python deeptrace.py run

# Run in headless mode
python deeptrace.py headless-run

# Get help
python deeptrace.py --help

Simple Web Interface

For quick testing, use the simplified Flask web app:

python simple_app.py

Then open http://localhost:5000 in your browser.

Optimizations

This project includes comprehensive optimizations for performance and efficiency:

Performance Enhancements

• ONNX Runtime Optimization: Leveraged FP16 precision and graph optimizations
• Multi-threading: Parallel processing for face detection and analysis
• GPU Acceleration: CUDA and DirectML support for faster inference
• Model Caching: Intelligent caching to reduce load times
• Memory Management: Efficient resource allocation and cleanup

Code Quality

• Python 3.13 Compatibility: Full support for latest Python features
• Error Handling: Robust exception handling and logging
• Type Hints: Complete type annotations for better IDE support
• Modular Architecture: Clean separation of concerns for maintainability

Benchmarks

Run performance tests:

python test_optimizations.py

For detailed optimization documentation, see OPTIMIZATION_COMPLETE.md and ADVANCED_OPTIMIZATIONS.md.

Documentation

FYP Documentation

• Technical Documentation - Deep dive into models and architectures
• Defense Guide - Q&A for FYP defense preparation
• Visual Diagrams - System architecture and data flow diagrams
• Quick Start Guide - Getting started with DeepTrace
• Migration Guide - Rebranding and optimization details

Features

• Face Detection: YOLO-Face, RetinaFace, SCRFD models
• Face Recognition: ArcFace (ResNet-50) for identity verification
• Face Swapping: InsWapper, GhostFace, BlendSwap, HyperSwap, SimSwap
• Face Enhancement: GFPGAN, Real-ESRGAN for quality improvement
• Lip Sync: Wav2Lip integration for talking video generation
• Audio Generation: Text-to-speech capabilities
• Modular Pipeline: Each component can run independently or chained together

Academic Use

This project is developed for academic research and FYP demonstration. Please use responsibly and ethically:

• Always watermark generated content
• Respect privacy and consent
• Follow institutional ethics guidelines
• Document all experiments and results

Repository

GitHub: https://github.com/Ali7040/Deepfake-gen-pipeline

License

See LICENSE.md for details.