Did you recently do ’random forest’ classification in SNAP with version 7 in windows?
It is ok for maximum likelihood but it is not working for ‘RF’ and classes are mixed.
It mixed classes with NAN that’s why I got training samples over ‘NAN’ area as well which previously worked but this time is not working.
Same problem like me? Any idea how we should do?
Actually I tested it for version 6 but same.
same problem here but no solusion
I am running a RF classification using multi-temporal S2 level 1C images (not processed to 2A yet). The classification looks good but some of the pixels are classified as NaN. Here are my steps:
-Resample (to 10 m)
-Subset (reduce size of images)
-Import Vector (shapefile of the study area)
-Reproject (to the same UTM zone of the original product. Necessary for the next step to work properly. No data is set to -1 and not NaN)
-Land-Sea mask to mask the images with the exact shape of th…
Have you made sure it is not related to the definition of the confidence threshold?
I have seen both cases in this forum:
pixels are classified as NaN because their confidence is too low
pixels are classified as NaN because one of the input bands is NaN at this location
@ABraun for reply…non of them…but I converted ‘NAN’ values to (-1000) and then it is solved now but I think this is a bug in the SNAP because sometimes I see this and everytime, I should find a new way to handle it.
I’m not sure if I understood correctly. You gave all NaN pixels in the classified product the value of -1000?
No, I gave all NAN values in input products ‘-1000’. My inputs values were between 0 to 255.
oh, I see.
Does it also work when the “Use NoData value” option is disabled in the band properties of all input rasters?