Question about converting Similarity3 to Pose3 in GTSAM (Monocular SLAM)

[tejalbarnwal]

Hi,
I am currently working on implementing a monocular SLAM prototype using GTSAM and had a question regarding the use of Similarity3 (Sim3) factors and their relationship with Pose3 (SE3).

Since monocular SLAM does not observe absolute scale, I am experimenting with Similarity3 transformations to account for the scale ambiguity in the optimization. After solving the graph, however, I would like to interpret the estimated states as Pose3 values.

My question is:
If the optimized transformation is a Similarity3, what is the recommended way to obtain the corresponding Pose3 representation?

For example, if the optimized Sim3 is:
rotation (R)
translation (t)
scale (s)

should the Pose3 translation simply be interpreted as (t), or should it be normalized (e.g., divided by the scale), or is there another standard approach within GTSAM for converting Sim3 results into Pose3 poses?