Original paper uses PRNet for depth map estimation

[himansh1314]

I read the CDCN paper and it uses PRNet for estimating the 32*32 depth map. However, here I didn't find anything as such. Can you please give a clarification on this? What exactly is the output? Is it a depth map? And if the output is depth map then how is the depth map provided as ground truth?

[voqtuyen]

The depth map here is just kind of pseudo map where all 1s represent a true face and all 0s represent a fake face