Camera calibration with ISAM2 incremental approach

[Smithangshu]

Hi,

First of all, thank you all who contributed to the development of this excellent library.

I am wondering if it is possible to find value of K while progressing with incremental approach with ISAM2. I have checked out the example of using GeneralSFMFactor2 for optimizing poses, landmarks and K. But in that we need to have some rough 3d estimation of landmarks. But in my case, at start we only have Pose, 2d observations of the landmarks and rough K.

There might be some possibilities of running ISAM2 based approach for first two poses or LOST triangulation after getting 2nd pose, feed result of landmarks to the GeneralSFMFactor2 approach. But I doubt this will not be an efficient approach especially for realtime application with low hardware capabilities like mobile devices.

Can you please suggest me the most optimum approach that I can take to proceed?