Question about BlendedMVS and training

[nobeldang]

Hi, thanks for sharing your great work. I am trying to write the dataset for the blendedmvs but I have two naive questions and if you can answer them it would help me.

1. Which BlendedMVS version you used for training? BlendedMVS, BlendedMVS+ OR BlendedMVS++?
2. The depth maps of BlendedMVS usually has some masked out regions and corresponding masked rgb images where values are all 0 in both the depth map and rgb image. How do you handle that or does the model ignores the regions where 0 value is encountered while training?

Example of masked regions:

Thanks!