Skip to content

Abhishekmishra2808/road-safety

BranchesTags

Repository files navigation

VisionX

AI-Powered Road Change Detection Platform

VisionX is an advanced computer vision system that compares two road surveillance videos (base and present) to automatically detect and analyze infrastructure changes, including new or removed traffic signs, road markings, potholes, and other road conditions.

VisionX Logo

🎯 Features

Core Capabilities

  • Dual Video Comparison - Upload base (reference) and present videos for intelligent comparison
  • Real-Time Detection - Live frame-by-frame analysis with WebSocket updates
  • Change Analytics - Comprehensive dashboard with charts, metrics, and insights
  • Infrastructure Focus - Detects traffic signs, road markings, potholes (excludes vehicles)
  • AI-Powered - Uses YOLOv8 deep learning model with custom detection algorithms

Analytics Dashboard

  • Key Metrics - Total additions, removals, and change severity assessment
  • Data Visualizations - Bar charts and pie charts for change distribution
  • Top Changes Table - Detailed breakdown of detected changes
  • Intelligent Insights - Auto-generated findings and recommendations
  • Export Results - Download complete analysis reports

🏗️ Architecture

Tech Stack

Backend (Python)

  • FastAPI - High-performance async web framework
  • Uvicorn - ASGI server with WebSocket support
  • PyTorch - Deep learning framework
  • Ultralytics YOLOv8 - Object detection model
  • OpenCV - Computer vision processing

Frontend (React)

  • React 18.2 - Modern UI framework
  • Vite - Fast build tool
  • Recharts - Data visualization library
  • Axios - HTTP client

Project Structure

road-safety/
├── api_server.py           # FastAPI backend server
├── comparison_worker.py    # Video comparison engine
├── detection_worker.py     # Object detection worker (optional)
├── yolov8n.pt             # YOLOv8 trained model
├── requirements.txt        # Python dependencies
├── frontend/              # React frontend application
│   ├── src/
│   │   ├── App.jsx        # Main application component
│   │   ├── ResultsAnalytics.jsx  # Analytics dashboard
│   │   ├── styles.css     # UJET-inspired minimalist design
│   │   └── main.jsx       # React entry point
│   ├── index.html
│   ├── package.json
│   └── vite.config.js
└── api_jobs/              # Temporary job storage (auto-created)

🚀 Getting Started

Prerequisites

  • Python 3.10 or higher
  • Node.js 16 or higher
  • CUDA 11.8 (optional, for GPU acceleration)
  • 8GB+ RAM recommended

Installation

1. Clone the Repository

git clone https://github.com/Abhishekmishra2808/road-safety.git
cd road-safety

2. Setup Python Environment

# Create virtual environment
python -m venv .venv

# Activate virtual environment
# On Windows:
.venv\Scripts\activate
# On Linux/Mac:
source .venv/bin/activate

# Install Python dependencies
pip install -r requirements.txt

3. Setup Frontend

cd frontend
npm install
cd ..

Running the Application

Start Backend Server (Terminal 1)

# Activate virtual environment first
.venv\Scripts\activate  # Windows
# source .venv/bin/activate  # Linux/Mac

# Start FastAPI server
uvicorn api_server:app --reload --port 8000 --host 0.0.0.0

Backend will be available at: http://localhost:8000

Start Frontend Server (Terminal 2)

cd frontend
npm run dev

Frontend will be available at: http://localhost:3000

Access the Application

Open your browser and navigate to: http://localhost:3000

📖 Usage Guide

Step 1: Upload Videos

  1. Select Base Video - Choose your reference/baseline road footage
  2. Select Present Video - Choose the current road footage to compare
  3. Click "START COMPARISON" button

Step 2: Monitor Progress

  • Watch real-time change detection feed
  • Track progress with live frame count and percentage
  • View quick stats of detected changes

Step 3: Analyze Results

Once comparison is complete, you'll see:

  • Key Metrics Dashboard - Total additions, removals, change severity
  • Bar Chart - Category-wise breakdown of changes
  • Pie Chart - Overall distribution of additions vs. removals
  • Top Changes Table - Detailed list of significant changes
  • Insights & Recommendations - AI-generated findings and action items

Step 4: Export Results

Click "DOWNLOAD RESULTS" to get the complete analysis report with:

  • Frame-by-frame comparison data
  • Annotated detection images
  • JSON data export

🔧 Configuration

Backend Settings

Edit api_server.py to customize:

  • PORT - Default: 8000
  • CORS_ORIGINS - Allowed frontend origins
  • UPLOAD_DIR - Job storage directory

Detection Settings

Edit comparison_worker.py to adjust:

  • CONFIDENCE_THRESHOLD - Default: 0.3
  • SKIP_FRAMES - Frame sampling rate (default: process 2 fps)
  • SPATIAL_THRESHOLD - Object matching distance (default: 50px)

Frontend Settings

Edit frontend/src/App.jsx to change:

  • API_URL - Backend server URL
  • Chart colors and styling
  • Severity thresholds

🎨 Design Philosophy

VisionX follows a minimalist design language inspired by UJET:

  • Clean black and white color palette
  • Bold typography with Inter font
  • Wide letter spacing for branding
  • Subtle animations and hover effects
  • Focus on content and functionality

🤖 Detection Capabilities

What VisionX Detects

✅ Traffic Signs - Stop signs, traffic lights, regulatory signs
✅ Road Markings - Lane lines, crosswalks, road paint
✅ Potholes - Road damage and surface defects
✅ Infrastructure - Poles, barriers, road furniture

What VisionX Ignores

❌ Vehicles (cars, trucks, buses, motorcycles)
❌ Boats
❌ Persons and bicycles

Focus is exclusively on static infrastructure and road conditions

📊 API Endpoints

POST /upload_comparison

Upload two videos for comparison

  • Parameters:
    • base_video (file) - Reference video
    • present_video (file) - Current video
  • Returns: { job_id: string }

WebSocket /ws/{job_id}

Real-time status updates

  • Sends: Progress, frame count, detected changes
  • Format: JSON messages

GET /results/{job_id}

Download comparison results

  • Returns: ZIP file with analysis data

🐛 Troubleshooting

Backend Won't Start

  • Ensure Python virtual environment is activated
  • Check if port 8000 is available
  • Verify all dependencies installed: pip list

Frontend Won't Start

  • Clear node_modules: rm -rf node_modules && npm install
  • Check if port 3000 is available
  • Verify Node.js version: node --version (should be 16+)

CUDA/GPU Issues

  • System will automatically fall back to CPU if CUDA unavailable
  • Check CUDA installation: python -c "import torch; print(torch.cuda.is_available())"
  • For CPU-only: Remove --index-url line from requirements.txt

WebSocket Connection Fails

  • Ensure backend is running on port 8000
  • Check CORS settings in api_server.py
  • Verify firewall isn't blocking WebSocket connections

🛠️ Development

Run in Development Mode

# Backend with auto-reload
uvicorn api_server:app --reload --port 8000

# Frontend with hot-reload
cd frontend && npm run dev

Build for Production

# Frontend production build
cd frontend
npm run build
# Output will be in frontend/dist/

# Serve frontend build
npm run preview

📝 License

This project is licensed under the MIT License.

👨‍💻 Author

Abhishek Mishra

🙏 Acknowledgments

  • Ultralytics - YOLOv8 object detection model
  • UJET - Design inspiration
  • FastAPI - Modern Python web framework
  • React - UI library

📈 Future Enhancements

  • Historical trend analysis
  • Multi-camera support
  • Custom object training
  • Email notifications
  • Cloud deployment guide
  • Mobile responsive improvements
  • Batch processing support

VisionX - Intelligent Infrastructure Monitoring

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

1 fork
Report repository

Releases

No releases published

Packages

No packages published

Languages