Hello everyone recently i am working on a assignment where i have to predict mask of image using foreground_background image,background image.I used bcewithlogitloss bcz i changed my target value as combination of 1 and 0 where 1 is for foreground and 0 is for background.So my results are pretty good but i am not still convinced why i used bcewithlogitloss.It is a reconstruction problem where i am using probabilty distribution(bcewithlogitloss).SO can anyone plz hepl me why i can use bcewithlogitloss in this example.Thank you