Closed collinjroland closed 3 days ago
Hi Colin,
Let me know how you go. It would also be worth checking the udm2 file in QGIS or similar just to double check that the Planet nan/cloud mask hasn't labelled a large portion of the image as cloud/nan. If that's not the case then it might be a bug that needs fixing in my codes,
Regards, Yarran
Description
When attempting to train a new classifier, calling
label_images
in theLabel images
cell oftrain_new_classifier.py
opens and then immediately closes the matplotlib window where labeling should occur, regardless of how high I set thecloud_thresh
.If I comment out lines 171-172 in
classifier_functions.py
, the labeling window remains open.This leads me to believe that there is an issue with the masks, probably something to do with the udm2 -> udm conversion or something else in the masking routine. Will continue to investigate.
Computing environment
OS: Windows 10 Enterprise Processor: 12th Gen Intel(R) Core(TM) i7-1265U
Conda environment