MambaSCD eval is NAN

[regainOWO]

Training Dataset: SECOND
Environment:

• Python: 3.12.3
• torch: 2.2.2+cu118

the first thing was GT_CD is no correct, but i found GT_CD data is not used to calculate eval score. my evaluation result is here

[ChenHongruixuan]

Hi,

May I ask what do you mean about [GT_CD is no corrent]? What is the value range of your generated binary change maps? The range of mine is 0-255, unchanged is 0 and changed is 255. please check this.

Best,

[ChenHongruixuan]

Also, the evalution code is from the following repo:
https://github.com/ggsDing/SCanNet
https://github.com/ggsDing/Bi-SRNet

[regainOWO]

The GT_CD that i set unchanged is 0 and changed is 1。and now i set GT_CD unchanged to 0 and changed to 255, it works fine!

thanks for your help!