A question about Mask

[Mrlong12]

what is the mask?

[libuyu]

You can refer to this closed issue: https://github.com/libuyu/GHM_Detection/issues/2

[Mrlong12]

Thanks

[Mrlong12]

And if I want to use the loss to train other task?such as the Semantic segmentation or saliency detection,the mask I should set 0.but how can i set the bin and the momentum value?

[libuyu]

If there is no region to ignore, your mask should be 1 everywhere. And each pixel in the final predicted map is a sample, and the pos/neg samples are not so imbalanced. So I don't think GHM is necessary.

If you want to use it, I think the bin can still be 10-40. But maybe you should try different momentum values.

[Mrlong12]

thanks , I will be try it.