How to apply scale and shift operations using noisy ground-truth depth?

Hello,

In Section 4 of your paper, you mentioned that noisy ground-truth depth can provide scale and shift parameters to align predicted relative depth maps with absolute depth measurements. Could you clarify the implementation details of this process? Specifically:

1. How are the scale and shift parameters derived from the noisy ground-truth depth?
2. What is the mathematical formulation for applying these parameters to the predicted depth?

A step-by-step description of the pipeline would be greatly appreciated. Thank you for your help!