RTMPose3d Z results interpretation

[BLukash]

Thanks @Tau-J for the awesome library, been using it for 2D keypoint inference with different versions of RTMPose2D for a long time now. As well as @S0yKaf for bringing 3d version here. However, I am curious on how to interpret Z values from the rtmpose3d onnx result. The following lines indicate that X and Y SimCC bins are transformed into pixel space of the original image dimensions, and Z is normalized into z_range, which almost mimics behavior from here with the exception of extracting z_root. As I understand @S0yKaf has exported model without built-int post-processing. My question is basically, can we transform Z (or X, Y) values to the coordinate space, where it can be used together with X and Y for proper 3D visualization and measurements.

Thank you in advance.