search

MarcoForte / FBA_Matting

Official repository for the paper F, B, Alpha Matting
MIT License
464 stars 95 forks source link

maybe a mistake about gradient exclusion loss? #33

Open wchstrife opened 3 years ago

wchstrife commented 3 years ago

In table 1, the gradient exclusion loss is calculated on the foreground and background of the original image, but i think is should be calculated on the predict foreground and background, Is this wrong?