Load Big GeoJSON Extension (QuPath 0.6.0)

v0.2.1 QuPath extension to import very large GeoJSON files using a streaming parser, so memory usage stays bounded while features are loaded.

What It Does

• Adds menu items under Objects:
  • Import large GeoJSON (streaming)...
  • Clear existing objects before import (toggle)
• Supports .geojson, .json, .geojson.gz, .json.gz
• Imports features in chunks (1000 by default)
• Converts supported GeoJSON geometry to QuPath ROI objects
• Creates:
  • cell objects when both geometry and nucleusGeometry are present
  • detection objects when properties.objectType == "detection"
  • annotation objects otherwise
• Sets object name from properties.name when present, otherwise from properties.cell_id

Supported Geometry Types

• Polygon
• MultiPolygon
• Point
• MultiPoint
• LineString
• MultiLineString

Build Requirements

• Java 21 (recommended for QuPath 0.6.0 compatibility)
• Gradle wrapper in this repo (./gradlew)

Build

./gradlew clean build

Built JAR:

build/libs/LoadBigGeoJSON-1.0.0.jar

Install In QuPath

1. Build the project.
2. Copy build/libs/LoadBigGeoJSON-1.0.0.jar into your QuPath extensions directory.
3. Restart QuPath.
4. Open an image and go to Objects menu.

Use From QuPath Menu

1. Open image in QuPath.
2. Optional: enable Clear existing objects before import.
3. Click Import large GeoJSON (streaming)....
4. Choose your GeoJSON file.
5. Check QuPath log for import summary.

Example Script

Annotated script:

scripts/import_large_geojson_via_extension.groovy

This script:

• opens a file chooser safely on JavaFX thread
• calls the extension importer by reflection
• lets you control chunkSize and clearExisting

GeoJSON Notes

• Root must be a JSON object containing a features array.
• Invalid/unsupported geometries are skipped.
• For polygons, rings are automatically closed if needed.

Troubleshooting

• No image is open; import canceled:
  • Open an image first, then rerun import.
• Dependency/build issues:
  • Use Java 21 and ./gradlew.
• Large files:
  • Keep chunk size at 1000 unless you have a reason to tune.