search

HUANGLIZI / LViT

[IEEE Transactions on Medical Imaging/TMI] This repo is the official implementation of "LViT: Language meets Vision Transformer in Medical Image Segmentation"
MIT License
298 stars 26 forks source link

关于n_labels #47

Closed xmeimeimei closed 4 months ago

xmeimeimei commented 4 months ago

您好 小白想请问一下既然最后分割应该是前景和背景两个类别的话 为什么n_labels也就是n_classes是1而不是2呢

HUANGLIZI commented 4 months ago

Because you only use 1 to indicate the foreground and use 0 in the same channel to indicate the background.