each instance in the label is assigned different pixel value?

[robbiewongBD]

according to the blob_loss.py，is each instance in the label image assigned different pixel value. For example, pixel 1 represent instance 1, pixel 2 represents instance 2. This is much different because we often label every instance within one class the same pixel value.

[neuronflow]

Thanks for your interest in blob loss. Yes, that is correct. In the paper, we use a connected component analysis to create these "multi" labels from binary label maps. There are many ways to implement this. My implementation is not really optimized for computation time.

[robbiewongBD]

would you like to provide the whole pipeline code of training a segmentation CNN using blob loss and making "multi" labels from binary label map.

[neuronflow]

I am currently focused on my PhD thesis. Once this is out of the way, I hopefully find time to do so.

[robbiewongBD]

hope you everything works out