I am working on a vegetation type classification using random forest and Sentinel-2 imagery, and I would like to export my final output map (GeoTIFF) which contains 7 classes and some masked areas with no value. However, when exporting the data to my gee assets with the pyramiding policies set to MODE and comparing the output to the original map, I find lots of discrepancies. Also, when I downloaded the output map to my computer I noticed that the masked areas were joined with the first class.
Am I exporting it wrong? Does anyone know what is happening and how I could fix that?
I basically have a trained classifier that I apply on a sentinel-2 mosaic with multiple bands and export it like this:
var trainedClassifier = ee.Classifier.smileRandomForest(100).train({
features: trainingSample,
classProperty: label,
inputProperties: bands
});
var imgClassified = INPUT_IMAGE.classify(trainedClassifier).updateMask(tree_mask);
var exportParams = {
image: imgClassified,
description: 'RF_my_map',
assetId: 'users/myuserID/name',
scale: 10,
region: geometry,
maxPixels: 2000000000
};
Export.image.toAsset(exportParams);
