matlabbe
matlabbe
In your tf tree, the base frame is `camera_link`. You can check if there is an offset between `camera_link` and `camera_color_optical_frame` with: ``` rosrun tf tf_echo camera_link camera_color_optical_frame ``` Normally...
How `map`->`camera_color_optical_frame` was recorded. With `frame_id` set to `camera_link`, you should see instead a TF tree like this: ``` map | -> odom | -> camera_link | -> camera_aligned_depth_to_color_frame |...
Could you remove `robot_localization` from the equation? to limit possible confusion with published TF frames. Because `robot_localization` localization would also publish `odom` or `map` frame at the same time than...
In blender, in which coordinates are you working? The TF pose would follow ROS convention, x->forward, y->left, z->up, while in blender it may be opengl coordinate frames: x->right, y->down and...
You cannot extract the pose like this. The transform with `camera_color_optical_frame` as child would be always `camera_aligned_depth_to_color_frame` from the tf tree you have show [here](https://github.com/introlab/rtabmap_ros/issues/993#issue-1795428662). The "pose" will be always...
Who is publishing that `map` -> `camera_color_optical_frame` ? Note also that you need to record `/tf_static` too in the bag.
rtabmap would publish `map`->`odom`, and visual odometry would publish `odom` -> `camera_color_optical_frame` **only** if `frame_id` for rtabmap is set to `frame_id`. I guess there is something else publishing `map` as...
Not in a tightly way, I think you may want to search for "visual inertial odometry" approaches, that could be integrated in rtabmap to replace its default one.
The last error is "out of GPU memory", you may need more memory. For special builds like you did, the long way is to uninstall `ros-$ROS_DISTRO-cv-bridge` completely, rebuild/install opencv with...
If the environment changed, there would be indeed less loop closures detected, or less inliers/matches between images of a loop closure. `rtabmap/localization_pose` is now always published and its covariance may...