About PPLC metric

[xiexh20]

Hi,

Thanks for open-sourcing the code, this is an amazing work!
In paper Section 3.4, you defined a new metric PPLC to evaluate the consistency between different views. This requires rectifying the second view to the first view. However, I could not find further details about this in the paper or in this repo. Could you explain this in more details? Namely,

1. How do you find the correspondences between to views to perform the rectification?
2. Do you assume the same camera intrinsics for all the methods at test time to do the rectification?

Thank you for your help!