Input data world-space vs camera-space?

[antithing]

Hi, and thanks for making this code available. I am trying to do bundle adjustment using fiducial markers found in two frames. I have couple of questions if you have a minute.

I have the marker corners as 3d points.

1 - Should my data into cvsba look like:

A- Marker corner positions in marker-space(distance from camera) and camera pose around zero. B- Marker corner positions in camera-space(around the zero point) and camera pose at distance from marker? C- somehow chain the markers together into world-space, along with the camera poses.(this is the result i want to get AFTER ba, do I need to do a rough version first?

2 - Is there a way to specify where I want the zero point of the resulting map to be?

Thanks again.

What I want to do is precisely map the fiducial markers in world space, using one of them as the zero point.