Description
Thanks for sharing the mapping tool!
I noticed that in the paper the ORB-SLAM2 package was utilized to generate camera tracking results, keyframe and associated landmarks which would help in building global-consistent map. While for the launch file you provided ('run_kitti.launch'), the tf_transform was used instead of SLAM poses, and I did not found the interfaces or topics subscribed associated with ORB-SLAM2. So would you plz guide me how the combine the wonderful mapping tool with state-of-the-art SLAM system which would output tracking results and pose graph optimization as well as loop closure information?
Meanwhile I ran the cblox mapping tool on the Euroc dataset rosbag which utilized dense stereo node of image_undistort package to generate pointclouds from stereo images, but the results were not as good as the original voxblox package. Below were the launch file and mesh pictures. The general TSDF server parameters were basically the same with voxel_size of 0.05 and pre_side of 32, so where I should revise to improve the mapping result?
