geobase/sparsemask

	GeoAi

task = land-cover-classification

🛠 Model Purpose

This model is part of the GeoAi javascript library.

GeoAi enables geospatial AI inference directly in the browser or Node.js without requiring a heavy backend.

GeoAi pipeline accepts geospatial polygons as input (in GeoJSON format) and outputs results as a GeoJSON FeatureCollection, ready for use with libraries like Leaflet and Mapbox GL.

---

🚀 Demo

Explore the model in action with the interactive Demo.

📦 Model Information

• Architecture: SparseMask
• Source Model: https://github.com/cliffbb/oem-lightweight
• Quantization: Yes

---

Note: This model only works if the input imagery contains both RGB and NIR bands.

💡 Example Usage

import { geoai } from "geoai";

// Example polygon (GeoJSON)
const polygon = {
  type: "Feature",
  properties: {},
  geometry: {
    coordinates: [
      [
        [-99.98170823334205, 50.64318249608178],
        [-99.9806018185983, 50.6431476272638],
        [-99.9807117895183, 50.642403469489295],
        [-99.98256386074388, 50.64247150726101],
        [-99.98170823334205, 50.64318249608178]
      ],
    ],
    type: "Polygon",
  },
} as GeoJSON.Feature;

// Initialize pipeline
const pipeline = await geoai.pipeline(
  [{ task: "land-cover-classification" }],
  providerParams
);
 
// Run detection
const result = await pipeline.inference({
  inputs: { polygon }
});
 
// Sample output format
// {
//     "detections": [
//      {type: 'FeatureCollection', features: Array(6)},
//      {type: 'FeatureCollection', features: Array(43)},
//      {type: 'FeatureCollection', features: Array(6)},
//      {type: 'FeatureCollection', features: Array(3)},
//      {type: 'FeatureCollection', features: Array(11)},
//      {type: 'FeatureCollection', features: Array(36)},
//      {type: 'FeatureCollection', features: Array(0)},
//      {type: 'FeatureCollection', features: Array(9)}
//    ]
//     "outputimage": GeoRawImage {data: Uint8ClampedArray(1048576), width: 512, height: 512, channels: 4, bounds: {…}, …}
// }

📖 Documentation & Demo

• GeoBase Docs: https://docs.geobase.app/geoai
• NPM Package: https://www.npmjs.com/package/geoai
• Demo Playground: https://docs.geobase.app/geoai-live/tasks/land-cover-classification
• GitHub Repo: https://github.com/decision-labs/geoai.js