enter image description here

I have performed random forest classification in Earth Engine would like to select only pixels with high probability. But I cant figure how it to be done. I see the probability value in the Inspector but cant figure out how to extract that value.

1 Answer 1


That number is not a probability. (The Inspector doesn't know anything about classification.) It is the mask of the image, which being less than 1 (100%) indicates that the particular pixel you clicked on was only partially covered by the original image data (or that some other mask manipulation occurred). The classifier does not care about mask; the mask value is the same at the output of the classifier as at the input.

  • Thanks for the reply, but I don't fully understand this. This particular pixel was in the middle of a farm in which all pixels were classified as '1'. I have performed crop mask and a NDVI mask, but the pixels removed due to masking are no where near this pixel. So what is the 'Original Image' here? Oct 11, 2022 at 5:50
  • @GeorgeBasil By "original image" I only mean "the input to the classifier". I can't say where it's coming from exactly — it would be necessary to experiment with the actual code to answer that. The one thing that I can tell you with certainty is that the % figure the Inspector displays is the mask value.
    – Kevin Reid
    Oct 11, 2022 at 14:30
  • Thank you. Can the probability value be extracted from a classified image in any way? Finding a few pixels with high confidence from the classification image is what I'm aiming for so that I can visit there and conduct some fieldwork. Oct 12, 2022 at 8:24
  • @GeorgeBasil Sorry, I'm not very familiar with classifiers. I believe some of them can be run in a probability mode, so that the band outputs would be probabilities instead of(?) a class, but I can't do better than "check the documentation" right now.
    – Kevin Reid
    Oct 12, 2022 at 14:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.