VisionCut - Background Removal Tool

This project provides a web application for removing backgrounds from images using various deep learning models. It features a Python FastAPI backend powered by PyTorch and a React frontend.

Features

• Background Removal: Upload images and remove their backgrounds automatically.
• Multiple Models: Choose from different background removal models (e.g., RMBG-2.0, BiRefNet).
• Model Refinement: Option to enable post-processing refinement for supported models.
• Web Interface: User-friendly React frontend for easy interaction.
• API: FastAPI backend provides endpoints for background removal and model management.
• CUDA Acceleration: Utilizes NVIDIA GPUs via PyTorch for faster processing if available.

Technology Stack

• Backend: Python, FastAPI, Uvicorn, PyTorch, Pillow
• Frontend: React, TypeScript, Material UI, Axios

Project Structure

. ├── downloaded_models/ # Default location for downloaded model files ├── frontend/ # React frontend application │ ├── public/ │ ├── src/ │ ├── package.json │ ├── Dockerfile # Frontend Docker build instructions │ └── docker-entrypoint.sh # Frontend container entrypoint script │ └── ... (other frontend files) ├── models/ # Model loading and registry logic │ ├── registry.py │ └── ... (model-specific files) ├── main.py # FastAPI application entry point ├── requirements.txt # Python dependencies ├── Dockerfile.backend # Backend Docker build instructions ├── docker-compose.yml # Docker Compose configuration ├── README.md # This file ├── run_backend.bat # Windows script to run backend locally ├── run_backend.sh # macOS/Linux script to run backend locally ├── run_frontend.bat # Windows script to run frontend locally ├── run_frontend.sh # macOS/Linux script to run frontend locally └── ... (other project files)

Setup

Prerequisites

• Python 3.8+ and Pip
• Node.js and npm (or yarn)
• (Optional but Recommended) NVIDIA GPU with CUDA drivers installed for GPU acceleration.

Backend Setup

1. Clone the repository:

   git clone <your-repository-url>
   cd pictransformer_birefnet

2. Create and activate a virtual environment (recommended):

   # Windows
   python -m venv venv
   .\venv\Scripts\activate
   
   # macOS/Linux
   python3 -m venv venv
   source venv/bin/activate

3. Install Python dependencies:

   pip install -r requirements.txt

   Note: Ensure your PyTorch version matches your CUDA setup if using GPU. You might need to install a specific PyTorch version from their website.

Frontend Setup

1. Navigate to the frontend directory:

   cd frontend

2. Install Node dependencies:

   npm install
   # or yarn install

3. (Optional) Configure API URLs: Create a .env.local file in the frontend directory if you need to override the default API endpoint (e.g., if your backend runs on a different address/port). See .env.development for examples.

   # frontend/.env.local
   REACT_APP_API_BASE_URL=http://localhost:8000
   REACT_APP_API_REMOVE_BG_URL=$REACT_APP_API_BASE_URL/remove-background
   REACT_APP_API_MODELS_URL=$REACT_APP_API_BASE_URL/models
   # ... add other URLs as needed

Running the Application

You can run the application locally using the provided scripts/manual steps, or use Docker Compose for a containerized setup.

Using Docker Compose (Recommended for Containerized Environment)

1. Prerequisites: Docker and Docker Compose installed.
2. Build and Run: From the project root directory, run:

   docker-compose up --build

   • This will build the images for both the backend and frontend (if not already built) and start the services.
   • The frontend will be accessible at http://localhost:3000.
   • The backend API will be accessible at http://localhost:8000 (though the frontend is configured to talk to it via the internal Docker network name backend).
3. Stopping: Press Ctrl+C in the terminal where docker-compose up is running. To remove the containers, run docker-compose down.

Using Local Run Scripts (for Development)

Using Run Scripts

• Windows:
  • Run Backend: run_backend.bat
  • Run Frontend: run_frontend.bat
• macOS/Linux:
  • Run Backend: chmod +x run_backend.sh && ./run_backend.sh
  • Run Frontend: chmod +x run_frontend.sh && ./run_frontend.sh

Manual Steps

1. Run the Backend Server:

   • Make sure you are in the project root directory (pictransformer_birefnet) and your virtual environment is activated.
   • Start the FastAPI server:

     uvicorn main:app --reload --host 0.0.0.0 --port 8000

   • The API will be available at http://localhost:8000.

2. Run the Frontend Development Server:

   • Navigate to the frontend directory in a separate terminal.
   • Start the React app:

     npm start
     # or yarn start

   • Open http://localhost:3000 in your browser.

API Endpoints

• POST /remove-background/: Upload an image file for background removal.
  • Query Parameters: model (string), enable_refinement (boolean)
• GET /models: List available models and their status.
• POST /models/{model_name}/load: Load a specific model.
• POST /models/{model_name}/unload: Unload a specific model.
• GET /models/{model_name}/info: Get details about a model.
• GET /health: Health check endpoint.

Model Management

Models are defined in the models/ directory and managed by models/registry.py. Models are downloaded on demand (or when explicitly loaded via the API) into the downloaded_models/ directory (by default).

Deployment