☀️ SolarScope

Illuminating the Future of Renewable Energy

Transforming satellite imagery into actionable solar energy insights using cutting-edge AI

🚀 Quick Start • 📖 Documentation • 🎯 Features • 🤝 Contributing

🌟 What is SolarScope?

SolarScope is an intelligent, AI-powered web application that revolutionizes solar energy assessment by automatically detecting:

🏗️ Buildings & Rooftops from satellite/aerial imagery
☀️ Existing Solar Panels on detected structures
📊 Installation Potential for renewable energy deployment

Using a sophisticated dual-model sequential pipeline, SolarScope combines DeepLab V3+ semantic segmentation with Faster R-CNN object detection to deliver precise, actionable insights for solar energy planning.

graph LR
    A[📤 Upload Image] -->|🔍| B[Building Detection]
    B -->|✅ Found| C[Solar Panel Detection]
    B -->|❌ None| D[Stop]
    C --> E[📊 Generate Report]
    style A fill:#4CAF50,stroke:#333,stroke-width:2px,color:#fff
    style B fill:#2196F3,stroke:#333,stroke-width:2px,color:#fff
    style C fill:#FF9800,stroke:#333,stroke-width:2px,color:#fff
    style E fill:#9C27B0,stroke:#333,stroke-width:2px,color:#fff

✨ Features

🤖 AI-Powered Detection DeepLab V3+ with ResNet-50 backbone Faster R-CNN for solar panel identification Sequential pipeline optimization Real-time confidence scoring	🎨 Modern Interface Drag & drop file upload Live progress indicators Interactive visualizations Responsive design
📊 Comprehensive Analytics Building coverage statistics Solar panel count & locations GeoJSON export capability Detailed confidence metrics	🚀 Production Ready RESTful API endpoints Health monitoring Error handling Scalable architecture

🏗️ Architecture

🔄 Detection Pipeline

┌──────────────┐
│ Upload Image │
│   📷 .tif    │
└──────┬───────┘
       │
       ▼
┌──────────────────┐
│ Building Detect  │
│  DeepLab V3+     │
└──────┬───────────┘
       │
       ├─────► Buildings Found? ──NO──► ❌ Stop
       │
       YES
       ▼
┌──────────────────┐
│ Solar Panel Det  │
│  Faster R-CNN    │
└──────┬───────────┘
       │
       ▼
┌──────────────────┐
│  📊 Results &    │
│   GeoJSON Export │
└──────────────────┘

🧠 AI Models

Model	Purpose	Architecture	Output
DeepLab V3+	Building Segmentation	ResNet-50 Backbone	Binary Mask (256×256)
Faster R-CNN	Solar Panel Detection	Object Detection	Bounding Boxes + GeoJSON

📁 Project Structure

SolarScope/
│
├── 🎯 app.py                      # Flask application server
├── 🧠 models_module.py            # AI model inference engine
├── ⚙️  config.py                   # Configuration & settings
├── 📋 requirements.txt            # Dependencies
├── 📖 README.md                   # Documentation (you are here!)
├── 🚫 .gitignore                  # Git exclusions
│
├── 🤖 models/                     # Pre-trained model weights
│   ├── deeplabv3plus_buildings_state_*.pth
│   └── best_model.pth
│
├── 📂 static/
│   ├── 📤 uploads/               # User uploaded images
│   └── 📊 results/               # Detection outputs & GeoJSON
│
├── 🎨 templates/
│   ├── index.html                # Main web interface
│   └── about.html                # About page
│
└── 📓 Deeplab V3+_train/         # Training notebooks
    └── deeplab_v3_building_detection.ipynb

🚀 Installation

Prerequisites: Python 3.8+ | pip | Git

📦 Quick Start

Windows (PowerShell)

# 1️⃣ Clone the repository
git clone https://github.com/Tushar-Surti/SolarScope.git
cd SolarScope

# 2️⃣ Create virtual environment
python -m venv venv
.\venv\Scripts\Activate.ps1

# 3️⃣ Install dependencies
pip install -r requirements.txt

# 4️⃣ Download pre-trained models
# Download from: https://drive.google.com/drive/folders/1Z471pir3Q7x096WscPsXnW3hgyMEmqk0
# Place in models/ directory

# 5️⃣ Verify installation
python config.py

# 6️⃣ Launch application
python app.py

Linux / macOS

# 1️⃣ Clone the repository
git clone https://github.com/Tushar-Surti/SolarScope.git
cd SolarScope

# 2️⃣ Create virtual environment
python3 -m venv venv
source venv/bin/activate

# 3️⃣ Install dependencies
pip install -r requirements.txt

# 4️⃣ Download pre-trained models (we have already trained it on our system,if you want to train it yourself you can run the training notebooks)
# Download from: https://drive.google.com/drive/folders/1Z471pir3Q7x096WscPsXnW3hgyMEmqk0
# Place in models/ directory

# 5️⃣ Verify installation
python config.py

# 6️⃣ Launch application
python app.py

🤖 Model Setup

# Download models from Google Drive
# https://drive.google.com/drive/folders/1Z471pir3Q7x096WscPsXnW3hgyMEmqk0

# Expected files in models/ directory:
# ├── deeplabv3plus_buildings_state_20250930_021413.pth
# └── best_model.pth

🎉 Ready to Go! Visit http://localhost:5000 in your browser

🎮 Usage Guide

🌐 Starting the Server

python app.py
# 🚀 Server running at http://localhost:5000

📸 Detection Workflow

1️⃣	Upload Image • Drag & drop or click to browse • Supported: PNG, JPG, TIFF, GeoTIFF • Recommended: GeoTIFF for geo-referencing
2️⃣	Configure Detection • Adjust confidence thresholds (optional) • Set window size & overlap for large images • Review detection parameters
3️⃣	Run Sequential Detection • Click "Start Sequential Detection" • Watch real-time progress indicators • Phase 1: Building detection → Phase 2: Solar panels
4️⃣	Analyze Results • View segmentation masks & overlays • Examine detected solar panels • Download GeoJSON files • Review statistics & metrics
5️⃣	Export & Share • Download visualization images • Export GeoJSON for GIS applications • Save TIFF masks for further analysis

📊 Detection Outputs

🏗️ Building Detection Results

Output	Description	Format
🎭 Segmentation Mask	Binary mask highlighting buildings	PNG
🔴 Overlay Image	Buildings highlighted in red on original	PNG
📈 Statistics	Coverage %, confidence, pixel counts	JSON
📐 Dimensions	Image resolution & processing details	JSON

☀️ Solar Panel Detection Results

Output	Description	Format
📦 Bounding Boxes	Detected panel locations with labels	PNG
🗺️ GeoJSON	Vectorized polygons with coordinates	GeoJSON
📋 Panel List	Individual detections with confidence	JSON
🎯 TIFF Masks	Geo-referenced detection masks	TIFF
📊 Count & Stats	Total panels & area coverage	JSON

📝 Example Output Structure

{
  "success": true,
  "building_detection": {
    "detected": true,
    "coverage_percentage": 32.5,
    "confidence": 0.92,
    "building_pixels": 84832
  },
  "solar_panel_detection": {
    "detected": true,
    "count": 12,
    "avg_confidence": 0.87,
    "geojson_path": "/static/results/solar_panels.geojson"
  }
}

⚙️ Configuration

Customize detection parameters in config.py:

# 🎯 Detection Thresholds
BUILDING_DETECTION_THRESHOLD = 1.0       # Minimum building coverage (%)
BUILDING_CONFIDENCE_THRESHOLD = 0.5      # Minimum confidence score
SOLAR_PANEL_CONFIDENCE_THRESHOLD = 0.5   # Panel detection threshold

# 🔧 Processing Parameters
SOLAR_PANEL_WINDOW_SIZE = 512            # Detection window size
SOLAR_PANEL_OVERLAP = 256                # Overlap between windows
MAX_VISUALIZATION_DIMENSION = 2048       # Max image size for display

# 🖥️ Server Settings
DEFAULT_PORT = 5000                      # Flask server port
DEBUG_MODE = True                        # Enable debug logging
DEVICE = 'cuda'                          # 'cuda' or 'cpu'

🐛 Troubleshooting

❌ Model Not Loading

# Verify model files exist
ls models/

# Check PyTorch installation
python -c "import torch; print(torch.__version__)"

# Ensure correct file names in config.py
# Expected: deeplabv3plus_buildings_state_*.pth and best_model.pth

⚠️ Solar Panel Detection Not Working

✅ Ensure input image is GeoTIFF format
✅ Verify geoai library is installed: pip show geoai
✅ Check model path in configuration
✅ Ensure buildings were detected in Phase 1

💾 Out of Memory Errors

# In config.py, reduce these values:
SOLAR_PANEL_WINDOW_SIZE = 256  # Default: 512
MAX_VISUALIZATION_DIMENSION = 1024  # Default: 2048

# Or force CPU usage:
DEVICE = 'cpu'  # Default: 'cuda'

📦 Import Errors

# Reinstall all dependencies
pip install -r requirements.txt --force-reinstall

# For geoai specifically
pip install geoai --upgrade

# Verify installations
pip list | grep -E "torch|flask|geoai"

🌐 Server Won't Start

Check port 5000 is not in use: netstat -ano | findstr :5000
Try different port in config.py: DEFAULT_PORT = 8000
Check firewall settings
Ensure virtual environment is activated

🔌 API Reference

`POST /detect`

Performs sequential building and solar panel detection.

Request:

POST /detect HTTP/1.1
Content-Type: multipart/form-data

file: <image_file>
confidence_threshold: 0.5 (optional)
window_size: 512 (optional)
overlap: 256 (optional)

Response:

{
  "success": true,
  "original_image": "/static/uploads/image.tif",
  "building_detection": {
    "detected": true,
    "prediction_mask": "/static/results/mask.png",
    "overlay_image": "/static/results/overlay.png",
    "statistics": {
      "coverage_percentage": 25.3,
      "confidence": 0.89
    }
  },
  "solar_panel_detection": {
    "detected": true,
    "count": 8,
    "detections": [...],
    "visualization": "/static/results/panels.png",
    "geojson": "/static/results/panels.geojson"
  }
}

`GET /health`

Health check endpoint for monitoring.

Response:

{
  "status": "healthy",
  "building_model_loaded": true,
  "solar_panel_model_loaded": true,
  "device": "cuda",
  "version": "1.0.0"
}

🔬 Technical Details

🏗️ Building Detection Model

Architecture: DeepLab V3+ with ResNet-50 backbone
Framework: PyTorch
Input Size: 256×256 RGB images
Output: Binary segmentation mask (building/background)
Training: Custom dataset with building annotations
Threshold: 1% building coverage to trigger solar detection
Preprocessing: Resize, normalize, tensor conversion

☀️ Solar Panel Detection Model

Architecture: Faster R-CNN
Framework: PyTorch with geoai wrapper
Input: GeoTIFF images (variable resolution)
Processing: Sliding window (512×512) with 256px overlap
Output: Bounding boxes + confidence scores + GeoJSON polygons
Post-processing: Non-maximum suppression, coordinate transformation
Geo-referencing: Maintains spatial coordinates in output

⚡ Performance Optimization

GPU Acceleration: CUDA support for faster inference
Batch Processing: Efficient window-based detection
Memory Management: Dynamic image resizing for large files
Caching: Model weights loaded once at startup
Asynchronous I/O: Non-blocking file operations

📄 License

This project combines functionalities from two separate detection systems. Ensure you have appropriate licenses for both models and the geoai library.

🤝 Contributing

We welcome contributions! Here's how you can help:

Code Improvements
Optimize algorithms
Add new features
Fix bugs

Documentation
Improve guides
Add tutorials
Write examples

Testing
Report issues
Test edge cases
Validate results

🎯 Areas for Improvement

🔋 Add support for batch processing multiple images
📊 Implement model versioning and A/B testing
🎨 Enhanced visualization with 3D building models
🗺️ Integration with mapping APIs (Google Maps, Mapbox)
📱 Mobile-responsive UI improvements
🔄 Real-time model retraining pipeline
📈 Advanced analytics dashboard
🌍 Multi-language support

📝 Contribution Guidelines

🍴 Fork the repository
🌿 Create a feature branch: git checkout -b feature/amazing-feature
💾 Commit changes: git commit -m 'Add amazing feature'
📤 Push to branch: git push origin feature/amazing-feature
🎉 Open a Pull Request

� License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

Built with ❤️ and powered by:

Special Thanks To

🎓 DeepLab V3+ - for semantic segmentation architecture
🔍 Faster R-CNN - for object detection framework
📦 geoai - Geospatial AI library for solar panel detection
🌍 PyTorch Team - For the amazing deep learning framework
🎨 Icons8 - For beautiful 3D icons

🌟 Star this repo if you find it useful! 🌟

Made with 💚 for a sustainable future

Empowering renewable energy decisions through artificial intelligence

⚠️ Important Notice

This project is for educational and research purposes only.
Commercial use or public deployment of this software requires explicit written permission from the author.
Unauthorized commercial or public use is strictly prohibited.

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
Deeplab V3+_train		Deeplab V3+_train
static		static
templates		templates
.gitignore		.gitignore
QUICKSTART.md		QUICKSTART.md
README.md		README.md
app.py		app.py
config.py		config.py
models_module.py		models_module.py
requirements.txt		requirements.txt
train_solar_panel_detection.ipynb		train_solar_panel_detection.ipynb

Folders and files

Latest commit

History

Repository files navigation

☀️ SolarScope

Illuminating the Future of Renewable Energy

🌟 What is SolarScope?

✨ Features

🤖 AI-Powered Detection

🎨 Modern Interface

📊 Comprehensive Analytics

🚀 Production Ready

🏗️ Architecture

🔄 Detection Pipeline

🧠 AI Models

📁 Project Structure

🚀 Installation

📦 Quick Start

🤖 Model Setup

🎮 Usage Guide

🌐 Starting the Server

📸 Detection Workflow

📊 Detection Outputs

🏗️ Building Detection Results

☀️ Solar Panel Detection Results

📝 Example Output Structure

⚙️ Configuration

🐛 Troubleshooting

🔌 API Reference

POST /detect

GET /health

🔬 Technical Details

📄 License

🤝 Contributing

🎯 Areas for Improvement

📝 Contribution Guidelines

� License

🙏 Acknowledgments

Special Thanks To

🌟 Star this repo if you find it useful! 🌟

⚠️ Important Notice

About

Topics

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

`POST /detect`

`GET /health`

Packages