search

YichiZhang98 / DTML

[PRCV 2021] The official code for "Dual-Task Mutual Learning for Semi-Supervised Medical Image Segmentation"
24 stars 2 forks source link

Help for understanding consistency loss #9

Closed ZhouCX117 closed 1 year ago

ZhouCX117 commented 1 year ago

Hi, thanks for sharing the code. dis_to_mask = torch.sigmoid(-1500*outputs_dis) in line 171 makes me confused. Why do you set the value to "1500"?

YichiZhang98 commented 1 year ago

k is a transform factor selected as large as possible to approximate the transform from distance maps to segmentation masks. Here we select k=1500.