I trained a 2-class model. For the background pixel (labeled in 0) , the prediction is ok. But for the foreground pixel, the prediction is 0.5 for label 0 and 0.5 for label 1.
No changes has made to the origin framework. Thanks for your reply.
prediction is double value like .49998, such value can never be strictly equal to .5
please explain clearly what you want, what you do and what happens instead
I trained a 2-class model. For the background pixel (labeled in 0) , the prediction is ok. But for the foreground pixel, the prediction is 0.5 for label 0 and 0.5 for label 1.
No changes has made to the origin framework. Thanks for your reply.