Adding custom odometry to RTABMAP

[Kavi91]

@matlabbe Sir, Please accept my sincere gratitude for implementing RTABMAP and making it open source for future developments.

So, I am planning to add my version of deep learning-based odometry into RTABMAP. This odometry mainly fused the camera and lidar together to produce a single odometry. Features will also extracted using Deep neural networks for both lidar and RGB.

In order to implement this with current framework.

1. Where I should place my new files

2. which files I should modify to enable custom feature extractions and odometry ?

Thank you in advance.

[Kavi91]

@matlabbe Sir, Please accept my sincere gratitude for implementing RTABMAP and making it open source for future developments.

So, I am planning to add my version of deep learning-based odometry into RTABMAP. This odometry mainly fused the camera and lidar together to produce a single odometry. Features will also extracted using Deep neural networks for both lidar and RGB.

In order to implement this with current framework.

1. Where I should place my new files
2. which files I should modify to enable custom feature extractions and odometry ?

Thank you in advance.

[matlabbe]

If you are using ROS, see comment here: https://github.com/introlab/rtabmap_ros/issues/1281

For the standalone version, that is more complicated. Basically, it would be to add a new class Odometry**** along those https://github.com/introlab/rtabmap/tree/master/corelib/src/odometry and override Odometry::computeTransform() abstract function. The SensorData object provided as argument would contain lidar and camera images + calibrations.

[Kavi91]

@matlabbe Thank you so much. I will try this.