Spatial Scene

A web application for processing images into 3D spatial scenes using depth estimation and inpainting techniques.

Features

• Depth Estimation: Generate depth maps from 2D images
• Image Inpainting: Fill gaps and enhance images using AI models
• 3D Visualization: Render layered depth images (LDI) with parallax effects
• Gallery Management: Save and manage processed scenes
• Cross-platform: Web-based interface with mobile support

Tech Stack

Backend

• FastAPI for REST API
• PyTorch for AI models
• Transformers for depth estimation
• OpenCV for image processing

Frontend

• React 18 with TypeScript
• Three.js with React Three Fiber
• Vite for development
• TailwindCSS for styling

Quick Start

Prerequisites

• Python 3.8+
• Node.js 18+
• Modern web browser

Setup

1. Backend Setup

cd backend
pip install -r requirements.txt
uvicorn main:app --reload --host 0.0.0.0 --port 8000

2. Frontend Setup

cd frontend
npm install
npm run dev

3. Access Application
   • Frontend: http://localhost:5173
   • Backend API: http://localhost:8000

Usage

1. Upload an image through the web interface
2. The system will automatically:
   • Generate a depth map
   • Create an inpainted version
   • Render as 3D layered scene
3. Interact with the 3D scene using mouse/touch
4. Save scenes to gallery for later viewing

Development

Use the provided scripts for easy development:

• ./start-dev.sh (Unix/Mac)
• start-dev.bat (Windows)

License

MIT License