would be useful to use CASENet for binary edge detection?

[NdaAzr]

Thank you for sharing your implementation.
I am wondering would be useful to use the CASENet network for the binary edge detection for the custom dataset or adopting the CASENet network with Hausdorff distance loss (or even cross-entropy loss) instead of multi-label loss?

Here is an example of my ground truth:

Thank you in advance

[anirudh-chakravarthy]

Hi,

Yes, I don't see why not. I think you could.