DeepScan API

DeepScan API is a FastAPI-based service that processes videos to detect deepfake content using deep learning models. It provides real-time analysis and visualization of the detection process.

Features

• Deep learning-based deepfake detection
• GradCAM visualization for model interpretability
• Efficient frame processing and storage
• REST API endpoints for video processing
• Integration with Laravel backend
• Configurable settings via YAML

System Requirements

• Python 3.8+
• CUDA-capable GPU (recommended)
• Storage space for processed frames

Installation

1. Clone the repository:

git clone https://github.com/recluzegeek/deepscan-api.git

cd deepscan-api

2. Create and activate a virtual environment:

python -m venv deepscan-api-venv
source deepscan-api-venv/bin/activate

OR

.\deepscan-api-venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure settings:

cp config/settings.yaml.example config/settings.yaml

Edit the settings.yaml file to configure the API host, port, and other parameters.

Project Structure

deepscan-api/
├── config/
│ ├── settings.example.yaml
│ └── settings.yaml
├── utils/
│ ├── model_manager.py
│ ├── database.py
│ ├── models.py
│ └── classification.py
├── services/
│ └── video_service.py
├── routers/
│ └── video_router.py
├── models/
│ └── weights/
├── requirements.txt
└── README.md

Usage

1. Run the API:

fastapi run main.py --reload --host 0.0.0.0 --port 8000

2. API Endpoints:

• POST /upload: Process video frames for deepfake detection

API Documentation

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

License

This project is licensed under the MIT License.