README.md

metadata

title: Elun
emoji: 🎨
colorFrom: blue
colorTo: purple
sdk: docker
pinned: false
license: mit

Elun - Image Processing Pipeline

A full-stack application that processes images through face parsing and upscaling using Real-ESRGAN and face-parsing.PyTorch models.

Project Structure

Elun/
├── backend/              # Python FastAPI server
│   ├── server.py        # Main FastAPI application
│   ├── config.py        # Configuration management
│   ├── main.py          # Alternative entry point
│   ├── requirements.txt  # Python dependencies
│   ├── face-parsing.PyTorch/  # Face parsing model
│   └── Real-ESRGAN/     # Image upscaler model
├── src/                 # React frontend
│   ├── App.tsx
│   └── components/
├── public/              # Static assets
├── vite.config.ts       # Vite configuration
├── package.json         # Node dependencies
└── tsconfig.json        # TypeScript config

Prerequisites

• Python 3.8+ (for backend)
• Node.js 16+ (for frontend)
• pip (Python package manager)
• npm or yarn (Node package manager)

Installation

1. Clone/Setup

Navigate to the project directory:

cd Elun

2. Backend Setup

Install Python dependencies:

cd backend
pip install -r requirements.txt

Note: The Real-ESRGAN and face-parsing models are large and may take time to download on first run.

3. Frontend Setup

Navigate back to root and install Node dependencies:

cd ..
npm install

Or with yarn:

yarn install

Running the Project

Option A: Using Batch Scripts (Windows)

From the project root, double-click:

• start-backend.bat - Starts the FastAPI server
• start-frontend.bat - Starts the Vite dev server

Option B: Manual Terminal Commands

Terminal 1 - Start Backend:

cd backend
uvicorn server:app --reload --port 5000

Terminal 2 - Start Frontend:

npm run dev

The frontend will be available at: http://localhost:5173

Configuration

Edit backend/config.py to adjust:

• Project paths
• Model settings
• Processing parameters

Project Features

• Image Upload: Upload images for processing
• Quality Options: Low, Medium, High (with Real-ESRGAN upscaling)
• Face Parsing: Automatically detects and processes facial regions
• Background Processing: Images process asynchronously in the background
• Real-time Updates: Frontend polls for processing status

API Endpoints

Endpoint	Method	Purpose
/process	POST	Upload image for processing
/health	GET	Health check

Next Steps

• Configure Supabase for image storage (optional)
• Add authentication
• Deploy to production
• Add more processing models

Troubleshooting

Backend won't start

• Check Python 3.8+ is installed: python --version
• Verify all dependencies: pip install -r requirements.txt
• Try running from the backend/ directory

Frontend won't start

• Clear node_modules: rm -r node_modules && npm install
• Check Node version: node --version

Models not downloading

• Ensure internet connection is stable
• Models download to backend/Real-ESRGAN/weights/ and backend/face-parsing.PyTorch/res/cp/
• First run may take 5-10 minutes

Development

• Frontend: React + TypeScript + Vite + Tailwind CSS
• Backend: FastAPI + Python subprocess for ML models
• ML Models: Real-ESRGAN (upscaling), face-parsing.PyTorch (segmentation)

License

See individual model repositories for licensing information.