Using a monocular camera to run rgbdslamv2

[zhangjd1029]

Hello, I deployed rgbdslamv2 on Ubuntu 20.04 and was able to successfully run the dataset. I tried using a monocular camera to obtain rgb images, using monodepth2 to obtain corresponding depth maps, and publishing/camera/rgb/colorimage, /camera/depth/image and /camera/rgb/camera Info, change the topic name in rgbdslam. launch, here is its rqt_ Graph

I have seen some examples of using an rgbd camera to run rgbdslamv2. After changing the topic name in rgbdslam. launch, they were able to run successfully. However, after doing so, the rgbdslam interface did not respond at all. May I know the reason for this? Looking forward to your! Thank you.

[zhangjd1029]

May I ask if the operation of rgbdslamv2 must include/tf topics？

[felixendres]

Hm, good question. rgbdslam_v2 relies on the topics being synchronized by _messagefilters (using message_filters::sync_policies::ApproximateTime). Maybe computing the depth image takes too much time?

Do you get the message from this line of code? with the appropriate topics?

If so, the next thing to check if you get callbacks to this callback method. Do you see its log messages?

[felixendres]

Tf shouldn't be required for the monocular case. You may see warnings, that a standard transformation from a base coordinate frame (Z points upwards) to a camera coordinate frame is assumed (where Z is forward).

[LiamHorton]

Hi @zhangjd1029:

I am also trying to deploy rgbdslamv2 on Ubuntu 20.04 with ROS noetic. I have been having issues with the PCL and OpenCV versions as well as the compiler option for PCL. I am currently using PCL 1.1 which requires C++14 or greater and OpenCV 4.2.

Could you provide what version of PCL and OpenCV you are using?

Thanks, Liam

[felixendres]

If you guys get it running, I'd be interested in a pull request. Please bear with me, if I don't respond timely though.