@tfzhou So, the class 0 is the ignore index and not the background class? I mean, for the background class is applied also the contrastive loss. Right?

[zhangyuxuan321]

          @tfzhou  So, the class 0 is the ignore index and not the background class? I mean, for the background class is applied also the contrastive loss. Right?

Originally posted by @nysp78 in https://github.com/tfzhou/ContrastiveSeg/issues/60#issuecomment-1287888818

I also have this confusion, and the number 0 also means the background class in my segmentation task. I can't figure out that whether to use it in author's contrastive method.

[tfzhou]

Hi @zhangyuxuan321, in my cases, class 0 indicates ignored categories. For cases that 0 refers to background, it should not be ignored for sure.

[zhangyuxuan321]

Thanks a lot for quick reply! Another question please, I would like to ask if there is available the implementation of sampling hard positive/negatives pixel for the computation of the contrastive loss, because I only found the implementation of hard anchors sampling.

Thanks

[heifanfanfan]

Hello, I would like to consult you about some problems related to code reproduction. If possible, can I get your contact information

[LePapilllon]

Hello, I want to know if I set ["reduce_zero_label" : false], then my background will not be ignored.