Scale Issues in Model Normalised Space

[Shexiaox]

I should like to enquire whether the scaling applied to the normalised space predicted by the model is consistent across both the camera pose's height and the translation transformation. Specifically, whether a single scale factor can be applied to restore the entire normalised space to its corresponding scale in the real world.

[Shexiaox]

Could there be discrepancies in the scale of the camera pose height and translation predicted by the model? If multiple cameras are predicted in a sequence, is the scale consistent for each prediction?