how to solve problem with Jacobians of different dimensions

[huntcheaters]

Hello,

I am working on visual odometry with hybrid error function.
e_rep is the error of reprojection whose Jacobian is a 26 matrix.
e_pho is the error of photometric whose Jacobian is a 16 matrix.
Maybe I have also other e_pho such as photometric error of depth image, e_dep.
I want to use a hybrid error e = e_rep + e_pho + e_dep.
To do this kind of optimization, how can I combine these Jacobians in one. I have only a single vertex which is the pose in SE3d. It is easy to calculate Jacobian of each type of error for a same vertext. By far, I have no idea to do that for a whole Jacobian.
BTW, I also want to give a weight of a single error, for which it is dependent on all errors. The computeError() computes only one error, so that it is not possible to give it a weight if I do not compute all errors before this. Is there a single way to do this.

Thanks.

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days.