FastFlowNet icon indicating copy to clipboard operation
FastFlowNet copied to clipboard

didn't process the occlusion loss?

Open poincarelee opened this issue 3 years ago • 1 comments

Hi, I have been trying to train fastflownet. After reference to the IRR-PWC, I found occlusion wasn't considered in fastflownet model, So I'd like to ask: if I want to use occlusion loss, how can occlusion be added to model part?

poincarelee avatar Mar 16 '22 06:03 poincarelee

For each decoder, you can change the output channels from 2 to 3, with one channel to estimate occlusion. Also, for the input of each decoder, you should add the occlusion map together with optical flow for coarse-to-fine estimation.

ltkong218 avatar Mar 18 '22 02:03 ltkong218