matlabbe

Is memory management enabled? If not, all nodes are in WM. Can you describe what issue you are trying to solve?

The natural way to make sure that the origin (or any important location of interest) is staying in WM is to keep the robot still when mapping to build up...

How did you timed it? Did you put a timer in python code around same function and print the result?

Interesting, it would mean that the python interpreter launched from c++ (we are now using https://github.com/pybind/pybind11 to manage python initialization and calling stuff) is slower than the one used when...

> There is no such difference. I also tried with hard coding the input data (loading it from disk in the python code) and not using the whole rtabmap framework...

Can you enable Debug log and show the log when a transformation is computed?

> Can one use pymatcher (e.g. SuperGlue) during localization, if KNN was used for mapping? Yes

`utility_export.h` should be [automatically generated on build](https://github.com/introlab/rtabmap/blob/afd10b1aa6f7727d9e19d6332b17e52e62468e43/utilite/src/CMakeLists.txt#L19-L20), which cmake version are you using?

[BEBLID](https://github.com/iago-suarez/BEBLID) is BSD license, so we could probably include their source code as 3rd party in rtabmap and use it when we detect OpencV

Having better distributed keypoints can increase robustness of visual odometry and better accuracy. The implementations seem not using the keypoint feature response to select "best" features (like the current [NMS](https://github.com/introlab/rtabmap/blob/e56e0c37f767987cd02407a0be28f0f40696768b/corelib/src/util2d.cpp#L2096-L2205)...