When calling seg_images_in_folder in Gym, the program calls the doodleverse utils fndo_seg:
The fn do_seg has a profile variable, which controls amount of output.
profile='minimal' is just the predicted segmentations
profile = 'meta' is all metadata and predicted segmentations
profile = 'full' is all metadata, predicted segmentations, probability maps, and overlays
full metadata is very heavy, esp. for many images, but the overlay and proability maps are so useful. I propose we add:
profile = 'overlay' for predicted segmentations, and overlays
profile = 'poverlay' for predicted segmentations, probability maps, and overlays
I can make this change in the coming days..
When calling
seg_images_in_folderin Gym, the program calls the doodleverse utils fndo_seg:doodleverse_utils/doodleverse_utils/prediction_imports.py
Line 296 in 69520b7
The fn
do_seghas aprofilevariable, which controls amount of output.profile='minimal'is just the predicted segmentationsprofile = 'meta'is all metadata and predicted segmentationsprofile = 'full'is all metadata, predicted segmentations, probability maps, and overlaysfull metadata is very heavy, esp. for many images, but the overlay and proability maps are so useful. I propose we add:
profile = 'overlay'for predicted segmentations, and overlaysprofile = 'poverlay'for predicted segmentations, probability maps, and overlaysI can make this change in the coming days..