wgcban / SemiCD

Revisiting Consistency Regularization for Semi-supervised Change Detection in Remote Sensing Images
https://www.wgcban.com/research#h.ar24vwqlm021
120 stars 19 forks source link

about the best_model.pth #5

Closed maozan closed 1 year ago

maozan commented 1 year ago

Excuse me, is the released model.pth file re-trained, it seems to be a little bit more accurate than what is reported in the paper.

wgcban commented 1 year ago

@maozan No, it is not re-trained - this is the model used to compute the metrics that are reported in the paper. Sometimes it may give slightly bad/good results depending on your GPU, and package versions.

maozan commented 1 year ago

@maozan No, it is not re-trained - this is the model used to compute the metrics that are reported in the paper. Sometimes it may give slightly bad/good results depending on your GPU, and package versions.

Thank you, the SemiCD is an impressive study. Looking forward to your more impactful achievements. @wgcban

wgcban commented 1 year ago

@maozan Thank you.