Having problems running AbsolutePoseSacProblem::TWOPT about opengv HOT 3 CLOSED

Hi Oliver,
I am trying to understand what's happening on your end, but I don't see an immediate solution. The first thing is that CentralAbsoluteAdapter is not necessarily something you want to use. It is very basic and has been designed mostly for the unit tests. If you use them, please verify that bearingVectors and points always have same length when you call the constructor. Also, TWOPT will be taking the rotation from the adapter as a guess to compute the translation. Note that you need to "load" this rotation before executing Ransac, and that there are currently no checks for the validity of the provided rotation.

from opengv.

Hi Laurent, thanks for your fast response.
I didn't see an immediate solution either;)
I am just doing basic things for the moment so I was pretty happy to use CentralAbsoluteAdapter. As I use the same datatypes, it seemed like a good fit.

The input bearings / points I give it are aligned, have the same length, and the bearings are normalised. I provide it a rotation estimate through the setR() method. The same input works 4 out of 5 times; but I don't see what causes that one other time to behave badly. The fact that it is returning the model size to be 0 sometimes is of concern - any idea how this could happen? Does it work for you if you change the algorithm to TWOPT in the unit test?

Cheers!

from opengv.

Problem solved by avoiding model-disambiguation for algorithm that compute a unique solution.

from opengv.