jianf-wang / np-semiseg Goto Github PK

Hello,

I am reaching out to inquire whether this method can be applied to image datasets that do not contain uncertainty labels. I noticed that the two datasets utilized in your paper both contain labels denoting uncertainty. I am interested to know if this approach can still be beneficial in contexts where such labels are not available.

Thank you for your time and consideration.

Hello, I'm reproducing a model using the VOC2012 dataset in AugSeg, and the test results I obtained have very confusing predicted labels. I can't even discern the rough shape of any object. What could be the reason for this?