Since < and > always return False with a NaN value, if a NaN is passed as a threshold, it'll result in a segmentation where each pixel is its own instance.
I don't think that feature can be useful to someone, moreover, the algorithm can take ages with a medium size image when we have that increasing number of segments to check for.
Since
<and>always return False with a NaN value, if a NaN is passed as athreshold, it'll result in a segmentation where each pixel is its own instance.I don't think that feature can be useful to someone, moreover, the algorithm can take ages with a medium size image when we have that increasing number of segments to check for.