Training of pairwise potentials of a dense CRF about dgm

[ereator]

Hi.
If I understand the question correctly, you wish to train the Bilateral Edge Model for the dense graph. You are right, that there is no training for that. (It is possible to train pairwise edge potentials, but they can't be used with the dense graph).

However, I think I can give you a hint how to tune the Bilateral Edge Model for your data. First of all, the Bilateral Edge Model is added to the graph via method void addBilateralEdgeModel(...). This method has a number of parameters and it is possible to train these parameters. Please see the Demo Param Estimation for the details how to do that.

Please let me know if this answers your question.

[I3aer]

Hi,

Thanks for your answer. I will check that demo to understand how to train parameters of bilateral kernels.