Drone-Based Road Damage Detection & Mapping

An AI-powered full-stack web application for automatic road crack detection and mapping using drone-captured images.

Features

• Upload drone images via a modern web dashboard
• Automatic crack segmentation using a pretrained U-Net (ResNet34) model
• GPS extraction from image EXIF metadata
• Overlay of detected cracks on the original image
• Interactive map with GPS marker and overlay preview
• Responsive, modern UI (React + TailwindCSS)

Tech Stack

• Frontend: React, Vite, TypeScript, TailwindCSS, Axios
• Backend: FastAPI, Python, segmentation-models-pytorch, OpenCV, Pillow, Folium, ExifRead
• Model: U-Net (ResNet34 encoder), trained on DeepCrack dataset

Project Structure

Drone Based/
├── backend/
│   ├── main.py
│   ├── utils.py
│   ├── model_loader.py
│   ├── models/
│   │   └── unet_deepcrack_best.pth
│   ├── static/
│   ├── maps/
│   ├── uploads/
│   └── requirements.txt
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   │   ├── UploadForm.tsx
│   │   │   ├── ResultPreview.tsx
│   │   │   └── MapViewer.tsx
│   │   ├── App.tsx
│   │   └── main.tsx
│   ├── index.html
│   ├── tailwind.config.js
│   └── package.json
└── README.md

Setup Instructions

1. Backend (FastAPI)

1. Install Python 3.8+ and pip.
2. Create and activate a virtual environment:

   cd backend
   python -m venv venv
   venv\Scripts\activate  # On Windows
   # or
   source venv/bin/activate  # On Mac/Linux

3. Install dependencies:

   pip install -r requirements.txt

4. Place your trained model at backend/models/unet_deepcrack_best.pth.
5. Start the backend server:

   uvicorn main:app --reload

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

2. Frontend (React + Vite)

1. Install Node.js (v18+) and npm
2. Install dependencies:

   cd frontend
   npm install

3. Start the frontend dev server:

   npm run dev

   The app will be available at http://localhost:5173.

Usage

1. Open http://localhost:5173 in your browser.
2. Upload a drone image (with GPS EXIF data).
3. View the detected cracks, GPS coordinates, and interactive map.

Notes

• The backend must be running for the frontend to work.
• Only images with GPS EXIF metadata will show a map location.
• For best results, use high-quality drone images.

Kaggle Reference

• Original U-Net Crack Detection Pipeline on Kaggle
• SUT-CRACK DATASET
• DEEPCRACK DATASET

License

This project is for academic/research use. Contact the author for other uses.