Depth L1 computation in evaluation

[JzHuai0108]

Dear Sir,

From your paper, I see that the depth L1 is used to evaluate the accuracy of rendered depth Dr relative to the ground truth depth Dg.
But Dg has a metric unit like m, and Dr is unitless because of the monocular setup.
How did you align up Dr and Dg in order to compute the depth L1 error?

I browsed the code base and found a bit clue in here.
But I am not sure about its meaning because this class is not used else where in the codebase.

Regards,