How do we calculate integer depth scale?

[shreyasaha1997]

I am trying to get instant ngp work with the co3d dataset, and I am able to get accurate results without using the depth inputs. I realize that I need to set the "integer_depth_scale" parameter in the transforms.json file in order to use the depth values and also saw in a couple of places that the default is 0.33. But I am not sure as to how it is calculated. Any help would be appreciated.