A comprehensive GUI application for automated photonic integrated circuit (PIC) layout verification using AI-powered discontinuity detection.

• GDS File Processing: Load and analyze GDS layout files
• SVG Conversion: Automatic conversion from GDS to SVG format
• Tile-Based Analysis: Split large layouts into manageable tiles for detailed inspection
• AI-Powered Detection: Uses Google's Gemini models for intelligent discontinuity detection
• Interactive Classification: Manual review and classification of detected issues

• Virtual Tile System: On-demand tile generation with intelligent caching
• Region of Interest (ROI): Focus analysis on specific layout areas
• Multi-threaded Processing: Configurable CPU usage for optimal performance
• Real-time Preview: Interactive navigation with coordinate transformation
• Comprehensive Export: JSON export with complete analysis metadata

• Dual-Panel Layout: Controls on left, original layout visualization on right
• Interactive Canvas: Click navigation, ROI drawing with blue boxes, and colored tile status overlays
• Visual Status Indicators: Color-coded dots (🟢/🔴/🟠) showing AI classification results
• Responsive Design: Components scale proportionally with window resizing
• Progress Tracking: Real-time analysis progress with pause/resume capability
• Enhanced Navigation: Previous/Next buttons with keyboard shortcuts for tile review

• Python 3.8+ (Recommended: Python 3.9 or higher)
• Operating System: macOS, Linux, or Windows
• Memory: 4GB RAM minimum, 8GB+ recommended for large layouts
• Storage: 1GB free space for temporary files and caching

• Google API Key: Required for AI analysis functionality
  • Get your key from Google AI Studio

One-command automated installation:

# Clone repository
git clone https://github.com/nguyennm1024/ai-photonic-physical-verification.git
cd ai-photonic-physical-verification

# Run automated installer (macOS, Ubuntu/Debian, Fedora)
./install.sh

# Setup your Google API key
./setup_api_key.sh

# Start the application
python3 main.py

That's it! The script automatically installs:

• Python 3.11 and dependencies
• Tcl/Tk GUI libraries
• SVG converter (librsvg)
• All Python packages

1. Operating System: macOS, Ubuntu/Debian, or Fedora
2. Google API Key: Get FREE key from Google AI Studio
3. 5-10 minutes for installation

If you prefer manual installation or use Windows:

# Clone and install
git clone https://github.com/nguyennm1024/ai-photonic-physical-verification.git
cd ai-photonic-physical-verification
pip install -e .

# Set API key
export GOOGLE_API_KEY='your_api_key_here'

# Install SVG converter
brew install librsvg  # macOS
# OR
sudo apt-get install librsvg2-bin  # Ubuntu/Debian

# Verify installation
python test_modules.py

Guides:

• 📖 First time? See QUICKSTART.md (3 simple steps!)
• 🔧 Detailed setup: See INSTALLATION.md
• ❓ Troubleshooting: See INSTALLATION.md#troubleshooting

python main.py

Note: The old layout_verification_app.py has been renamed to layout_verification_app_OLD.py. Use main.py for the refactored version.

1. Load GDS File: Click "Load GDS File" and select your layout file
2. Configure Grid: Set tile grid size (default 50×50), overlap (default 0%)
3. Generate Grid: Click "Generate Grid" to prepare the tile system
4. Select Region (Optional):
   • Click "Select ROIs" to enable ROI mode (button turns to "✓ Selecting ROIs")
   • Draw blue boxes on the layout to select specific regions
   • Select multiple regions without re-clicking the button
5. Start AI Analysis:
   • Click "Process All Tiles" for full layout analysis
   • Or "Process Selected Regions" for ROI-only analysis
6. Review Results:
   • View colored tile overlays on layout (green/red boxes)
   • Click any tile to see details in the review panel
   • Status indicators show 🟢 Continuity, 🔴 Discontinuity, or 🟠 No waveguide
   • Use Previous/Next buttons to navigate through tiles
7. Manual Classification: Classify tiles as "✓ Continuous" or "⚠ Discontinuity"
8. Export Results: Save analysis results to JSON format

Tips:

• Use ROI selection for focused analysis of critical areas
• Click "Cancel Processing" to pause at any time
• Resize window to adjust UI layout proportionally

Modular Structure:

• core/ - Business logic (file management, tile system, AI analyzer, ROI manager, state)
• utils/ - Coordinate transformations, image processing, thread-safe communication
• ui/ - Reusable components, event handlers, styling

The clean interface showing the dual-panel layout with controls on the left and interactive layout visualization on the right.

• INSTALLATION.md - Complete installation guide with troubleshooting
• docs/REFACTORING_DESIGN.md - Architecture & design details

Dual Model Architecture:

• Gemini Pro for detailed analysis and reasoning
• Gemini Flash for fast three-category classification

Analysis Pipeline:

1. Tile Generation: Create analysis-ready image tiles with resolution-aware caching
2. AI Processing: Send tiles to Gemini Pro for detailed analysis
3. Classification: Three-category classification using Gemini Flash:
   • Continuity - Waveguides are continuous and properly aligned
   • Discontinuity - Breaks, gaps, misalignments, or structural issues detected
   • No Waveguide - No actual waveguides found in the tile
4. Visual Feedback: Colored overlays on layout and status indicators in review panel
5. Result Storage: Store detailed analysis, classification, and coordinates
6. User Review: Manual verification and correction with Previous/Next navigation

Common Issues:

• Installation errors → See INSTALLATION.md
• Tile generation fails → Install rsvg-convert or inkscape
• Memory issues → Reduce grid size, use virtual tiles
• Analysis errors → Check API key configuration

# Configuration for quick analysis
Grid Size: 10×10
Tile Resolution: 512px (AI), 384px (preview)
CPU Cores: 4
Virtual Tiles: Enabled
ROI Analysis: Optional - select specific areas

# Default configuration - optimal for most cases
Grid Size: 50×50
Tile Overlap: 0%
Tile Resolution: 512px (AI), 384px (preview)
CPU Cores: 8
Cache Size: 100 tiles
Virtual Tiles: Enabled

# Configuration for comprehensive analysis
Grid Size: 80×80
Tile Resolution: 1024px
CPU Cores: 8
Virtual Tiles: Enabled
ROI Analysis: Highly recommended - focus on critical areas first
Preview Resolution: 384px for faster navigation

We welcome contributions! Here are some areas where you can help:

• Performance optimization for large layouts
• Additional AI models integration
• Export format extensions
• UI/UX improvements
• Documentation enhancements

1. Fork the repository and create a feature branch
2. Follow INSTALLATION.md for setup
3. Review docs/REFACTORING_DESIGN.md for architecture
4. Make your changes with tests
5. Submit a pull request

This project is open source. Please see the license file for details.

• GitHub Repository
• Google AI Studio - Get API Key
• Report Issues

Current Version: 2.1.0 Status: Active Development Last Updated: October 2025

v2.1 (October 2025) - UI/UX Enhancements:

• Visual status indicators (🟢/🔴/🟠), responsive design, enhanced ROI selection
• Accurate coordinate transformation, resolution-aware caching
• Three-category classification, updated defaults (50×50 grid)

v2.0 (October 2025) - Major Refactoring:

• Modular architecture with 24+ modules, comprehensive testing
• Validated with real 4.5MB GDS files

v1.0 (August 2025) - Initial Release:

• Complete GUI with AI-powered discontinuity detection