Personal Initiative AI · Open Source

Indoor Mapping MCP

Automating IMDF for Google Content Partners

Indoor Mapping Architecture Diagram

Empowering New Spatial Experiences

This project is a personal initiative designed to dramatically lower the barrier of entry for spatial data creation. By automating the IMDF pipeline, it serves as a foundational tool to build new businesses and applications around Indoor Mapping, Wayfinding, and Immersive View experiences.

The Context & Problem

Bringing large venues (stadiums, malls, airports) into the digital world requires strict compliance with the Indoor Maps Data Format (IMDF). For Google Maps Content Partners (GMCP), this process was historically a massive bottleneck.

The Old Way

  • Manual Tracing: GIS engineers manually georeference PDF floor plans.
  • Slow Process: Takes 2 to 3 days per venue to map.
  • Error-Prone: Manually generating 16 GeoJSON layers leads to structural mistakes.
  • High Rejection Rate: Failing any of the 50+ Google Maps schema rules restarts the process.

The MCP Solution

  • Automated Extraction: Node.js and Python pull geometries directly from PDFs.
  • Lightning Fast: Fully packaged ZIP delivered in ~15 minutes.
  • Built-in Validator: AI agent cross-references relationships and alignments automatically.
  • Zero Rejections: Actionable warnings fix issues before submitting to Google.

How the AI Pipeline Works

I built a unified Model Context Protocol (MCP) server that encapsulates the entire IMDF workflow into 5 deterministic steps. This allows AI assistants (like Claude) or automated scripts to handle geospatial reasoning without human intervention.

1. Acquire FootprintGoogle Places + OSM → GeoJSON
2. Extract FloorsPDF → 16 IMDF layers, georeferenced
3. Write LayersManual patch or inspect layers
4. Validate IMDFGMCP schema + geometry checks
5. Package ZIPDelivery-ready file for Google Maps

Example Workflow via MCP

// 1. Fetch real-world coordinates from address acquire_footprint({ query: "Caesars Superdome, New Orleans, LA" }) // 2. Extract architectural floors and anchor them to the footprint extract_floor_plan({ pdf_path: "floorplan.pdf", venue_footprint_geojson: {...}, level_names: ["Ground","Loge","Suite"] }) // 3. Ensure 100% compliance with Google Maps requirements validate_imdf_for_google({ dataset_path: "/output/superdome_imdf", profile: "google_maps", fail_on: "error" }) // 4. Create the submission package package_imdf_zip({ dataset_path: "/output/superdome_imdf", output_zip_path: "Superdome_IMDF.zip" })

Results & Business Impact

100%
GMCP Pass Rate
~15m
Per Venue (from days)
16/16
Layers Auto-Generated

By transforming a manual GIS task into a programmatic API, this open-source tool allows startups and agencies to scale indoor mapping projects rapidly. It clears the path for developing advanced wayfinding apps, seamless indoor-outdoor navigation, and rich immersive views for retail and real estate.

Explore the Code on GitHub

Next Steps & Roadmap

  • Support for multi‑building campuses and improved outdoor-indoor transition layers.
  • Integration with direct Google Maps Content Partners upload API for push‑button submission.
  • Additional AI‑powered floor plan cleanup using computer vision to correct OCR artifacts from old architectural drawings.
Dashboard or Immersive View Concept

Example of an immersive view output enabled by automated IMDF.