Export.classifier.toAsset

AI-generated Key Takeaways

• The Export.classifier.toAsset function creates a batch task to export an ee.Classifier as an Earth Engine asset.

• This function supports exporting ee.Classifier.smileRandomForest, ee.Classifier.smileCart, ee.Classifier.DecisionTree, and ee.Classifier.DecisionTreeEnsemble classifiers.

• The function requires specifying the classifier to export, and optionally allows setting a description, assetId, and priority for the task.

• Example code is provided for both JavaScript (Code Editor) and Python (Colab) to demonstrate how to use Export.classifier.toAsset with a trained smileRandomForest classifier.

• The examples show the process of gathering training data, training a classifier, creating the export task, and then loading and visualizing the exported classifier asset.

Creates a batch task to export an ee.Classifier as an Earth Engine asset.

Only supported for ee.Classifier.smileRandomForest, ee.Classifier.smileCart, ee.Classifier.DecisionTree and ee.Classifier.DecisionTreeEnsemble.

Usage	Returns
Export.classifier.toAsset(classifier, description, assetId, priority)

Argument	Type	Details
classifier	ComputedObject	The classifier to export.
description	String, optional	A human-readable name of the task. Defaults to "myExportClassifierTask".
assetId	String, optional	The destination asset ID.
priority	Number, optional	The priority of the task within the project. Higher priority tasks are scheduled sooner. Must be an integer between 0 and 9999. Defaults to 100.

Examples

Code Editor (JavaScript)

// First gather the training data for a random forest classifier.
// Let's use MCD12Q1 yearly landcover for the labels.
var landcover = ee.ImageCollection('MODIS/061/MCD12Q1')
    .filterDate('2022-01-01', '2022-12-31')
    .first()
    .select('LC_Type1');
// A region of interest for training our classifier.
var region = ee.Geometry.BBox(17.33, 36.07, 26.13, 43.28);

// Training features will be based on a Landsat 8 composite.
var l8 = ee.ImageCollection('LANDSAT/LC08/C02/T1')
  .filterBounds(region)
    .filterDate('2022-01-01', '2023-01-01');

// Draw the Landsat composite, visualizing true color bands.
var landsatComposite = ee.Algorithms.Landsat.simpleComposite({
  collection: l8,
  asFloat: true
});
Map.addLayer(landsatComposite, {
  min: 0,
  max: 0.3,
  bands: ['B3', 'B2', 'B1']
}, 'Landsat composite');

// Make a training dataset by sampling the stacked images.
var training = landcover.addBands(landsatComposite).sample({
  region: region,
  scale: 30,
  // With export to Classifier we can bump this higher to say 10,000.
  numPixels: 1000
});

var classifier = ee.Classifier.smileRandomForest({
  // We can also increase the number of trees higher to ~100 if needed.
  numberOfTrees: 3
}).train({features: training, classProperty: 'LC_Type1'});

// Create an export classifier task to run.
var assetId = 'projects/<project-name>/assets/<asset-name>';  // <> modify these
Export.classifier.toAsset({
  classifier: classifier,
  description: 'classifier_export',
  assetId: assetId
});

// Load the classifier after the export finishes and visualize.
var savedClassifier = ee.Classifier.load(assetId)
var landcoverPalette = '05450a,086a10,54a708,78d203,009900,c6b044,dcd159,' +
  'dade48,fbff13,b6ff05,27ff87,c24f44,a5a5a5,ff6d4c,69fff8,f9ffa4,1c0dff';
var landcoverVisualization = {
  palette: landcoverPalette,
  min: 0,
  max: 16,
  format: 'png'
};
Map.addLayer(
    landsatComposite.classify(savedClassifier),
    landcoverVisualization,
    'Upsampled landcover, saved');

Python setup

See the Python Environment page for information on the Python API and using geemap for interactive development.

import ee
import geemap.core as geemap

Colab (Python)

# First gather the training data for a random forest classifier.
# Let's use MCD12Q1 yearly landcover for the labels.
landcover = (ee.ImageCollection('MODIS/061/MCD12Q1')
             .filterDate('2022-01-01', '2022-12-31')
             .first()
             .select('LC_Type1'))

# A region of interest for training our classifier.
region = ee.Geometry.BBox(17.33, 36.07, 26.13, 43.28)

# Training features will be based on a Landsat 8 composite.
l8 = (ee.ImageCollection('LANDSAT/LC08/C02/T1')
      .filterBounds(region)
      .filterDate('2022-01-01', '2023-01-01'))

# Draw the Landsat composite, visualizing true color bands.
landsatComposite = ee.Algorithms.Landsat.simpleComposite(
    collection=l8, asFloat=True)

Map = geemap.Map()
Map  # Render the map in the notebook.
Map.addLayer(landsatComposite, {
    'min': 0,
    'max': 0.3,
    'bands': ['B3', 'B2', 'B1']
}, 'Landsat composite')

# Make a training dataset by sampling the stacked images.
training = landcover.addBands(landsatComposite).sample(
    region=region,
    scale=30,
    # With export to Classifier we can bump this higher to say 10,000.
    numPixels=1000
)

# We can also increase the number of trees higher to ~100 if needed.
classifier = ee.Classifier.smileRandomForest(
    numberOfTrees=3).train(features=training, classProperty='LC_Type1')

# Create an export classifier task to run.
asset_id = 'projects/<project-name>/assets/<asset-name>'  # <> modify these
ee.batch.Export.classifier.toAsset(
    classifier=classifier,
    description='classifier_export',
    assetId=asset_id
)

# Load the classifier after the export finishes and visualize.
savedClassifier = ee.Classifier.load(asset_id)
landcover_palette = [
    '05450a', '086a10', '54a708', '78d203', '009900',
    'c6b044', 'dcd159', 'dade48', 'fbff13', 'b6ff05',
    '27ff87', 'c24f44', 'a5a5a5', 'ff6d4c', '69fff8',
    'f9ffa4', '1c0dff']
landcoverVisualization = {
    'palette': landcover_palette,
    'min': 0,
    'max': 16,
    'format': 'png'
}
Map.addLayer(
    landsatComposite.classify(savedClassifier),
    landcoverVisualization,
    'Upsampled landcover, saved')