search

meng-tang / rloss

Regularized Losses (rloss) for Weakly-supervised CNN Segmentation
MIT License
208 stars 46 forks source link

can the densecrf loss use for medical image(1 channel)? #5

Open brisyramshere opened 5 years ago

meng-tang commented 5 years ago

The current implementation only supports 3 channels. However, you can duplicate the image to give an input blob of 3HW.

brisyramshere commented 5 years ago

The current implementation only supports 3 channels. However, you can duplicate the image to give an input blob of 3_H_W.

is the output of denseCRFLoss negative?and how to choose the weight of denseCRFLoss?

firelgh commented 4 years ago

The current implementation only supports 3 channels. However, you can duplicate the image to give an input blob of 3_H_W.

is the output of denseCRFLoss negative?and how to choose the weight of denseCRFLoss?

+1 I have the same question! is the output of denseCRFLoss negative?

convnets commented 4 years ago

The current implementation only supports 3 channels. However, you can duplicate the image to give an input blob of 3_H_W.

is the output of denseCRFLoss negative?and how to choose the weight of denseCRFLoss?

+1 I have the same question! is the output of denseCRFLoss negative?

+1 Not sure if the CRFLoss is negative. In my case, it is negative.