Kyle Barron - Lonboard: Fast, interactive geospatial vector data visualization in Jupyter

Discover Lonboard: a Python library revolutionizing Jupyter geospatial visualization with GPU acceleration, handling millions of points efficiently via GeoParquet and GeoArrow.

Key takeaways
  • Lonboard is a Python library for fast, interactive geospatial vector data visualization in Jupyter, capable of handling millions of points efficiently

  • Uses GeoParquet for efficient data transfer to browser and GeoArrow for binary data handling, making it significantly faster than traditional libraries like iPyleaflet

  • Can render up to 15 million points on capable hardware, with typical performance of 3 million points in about 2 seconds

  • Provides integration with popular tools like Panel, Jupyter widgets, and MapLibre for base maps

  • Features interactive GPU-accelerated visualizations with support for:

    • Point data
    • Line strings
    • Polygons
    • Animations
    • Custom coloring and styling
  • Allows export to standalone HTML files that contain all data and can be shared easily

  • Supports multiple data sources including:

    • GeoPandas DataFrames
    • DuckDB queries
    • Shapefiles
    • Any GDAL-supported format
  • Includes a CLI for quick visualization of geospatial data files

  • Uses Deck.GL under the hood for GPU-accelerated rendering in the browser

  • Focuses on being fast, simple to use, and interactive within the Jupyter ecosystem