New version of realsense for pointcloud it is hanging rviz #1886
Comments
|
Hi @ranjitkathiriya Each RealSense ROS wrapper version should be matched with a specific librealsense SDK version recommended in the wrapper release notes for that version. For example, RealSense ROS wrapper 2.2.24 and librealsense SDK 2.44.0. https://github.com/IntelRealSense/realsense-ros/releases When updating the wrapper version, have you first installed the recommended librealsense version, and then secondy built the ROS wrapper once librealsense has been successfully installed? Every time that librealsense is updated to a new version, the ROS wrapper should be built again afterwards. |
|
How should I completely remove LibRealSense v2.36.0? |
|
What this message is basically saying is that now that you have installed librealsense 2.45.0, build ROS wrapper 2.3.0 again and then this message should disappear. This is because if you build the wrapper with a particular librealsense version and then update librealsense, the wrapper has to be built again afterwards or this "built with, running with" message occurs. |
|
Yes I have done build things with 2.3.0
Then also I am facing the same issue.. |
|
You mean that it is still stating Built with LibRealSense v2.36.0, Running with LibRealSense v2.45.0 even though you have rebuilt the wrapper? |
|
Yes, |
|
Thank you. And was librealsense installed with Debian packages or built from source code? |
I have followed these steps. |
|
You have installed from Debian packages then. A RealSense team member provides advice in the link below for uninstalling librealsense packages. IntelRealSense/librealsense#7052 (comment) |
|
I have run into the same issue after updating, and have followed the instructions in your last comment.
|
|
Hi @jbeck28 Your version details look correct, as both Built With and Running With have the same librealsense version number. Your RealSense ROS version of 2.3.0 is also the correct one for use with 2.45.0. |
|
That is promising. I however, still have the issue of slow frame-rates, and when using a simple chessboardfinder, there are NaN's now where previously there were not. Any other ideas on what may resolve these problems? |
|
@jbeck28 In case investigation of your problems turns out to be complex, I recommend creating a new issue by visiting the link below and clicking the green New Issue button. Please include in the message which camera model you are using. Thanks! https://github.com/IntelRealSense/realsense-ros/issues |
|
Hello @MartyG-RealSense , I have followed your steps then also I am getting this warning. Is there any way then I can manually delete the file or I can change the build path? |
|
Hi @ranjitkathiriya Have you tried the purge and install method that I suggested yesterday, please? |
|
YesI tried this all command and tried |
|
Yes, Solved by removing all build, devel, logs folder and building catkin_ws again. Now it is solved. Other things I am getting this warning it is normal? What should be the minimum FPS in rviz again I am getting around 4 or 5.
And In real sense viewer, this is my frames drop per second graphs.
|
|
I would not recommend enabling RGB, Depth and Motion Module all at the same time in the RealSense Viewer. If you use 3 streams, when the Motion Module is enabled then it can cause the RGB stream to become No Frames Received. It does not matter which order the streams are enabled in, because if the problem occurs then the trigger for it is always the enabling of the Motion Module. So I would suggest only enabling RGB and Depth at the same time if possible. There is no fix that can correct the problem if it happens to you except for rebooting the computer. Your FPS problem in RViz sounds like the one jbeck28 was experiencing with rs_rgbd.launch. |
|
I am experiencing this command. roslaunch realsense2_camera rs_rgbd.launch filters:=colorizer,pointcloud align_depth:=true I have restarted my pc I have a clear catch and everything I have tried. Then also I am not able to fix this issue. In RVIZ I am getting fps in between 2-8. |
|
What kind of point cloud do you get if you remove align_depth:=true from the roslaunch instruction? |
|
if an organized cloud has invalid points and so is_dense = false then do you need to include is_dense:=false in the roslaunch instruction for roslaunch realsense2_camera rs_rgbd.launch filters:=colorizer,pointcloud ordered_pc:=true - Organized is_dense=False ? Also, if you are now able to set ordered_pc:=true from wrapper version 2.2.22 onward then could you instead use roslaunch realsense2_camera rs_camera and set the default unordered cloud in the rs_camera mode to ordered, if the need for an ordered cloud is the main reason for you using rs_rgbd |
|
I am sorry for the mistake I have updated the question again. |
And When I run this command:
What I wanted for my project: |
|
The pointcloud example in the RealSense ROS documentation seems to map RGB color onto depth on a pointcloud pretty well whilst only using roslaunch realsense2_camera rs_camera.launch filters:=pointcloud https://github.com/IntelRealSense/realsense-ros#point-cloud
|
Yes, This works perfectly for me even I am getting 28-31 fps with this command.
Over here I am calculating the pose, angle, and centre point for a particular object from coloured depth stream.
You can see the green point in the second Image. It is calculated with the help of colourized depth stream. My question was: is there any way that I can get point cloud much faster with rgb depth stream? |
|
The warning No stream match for pointcloud chosen texture Process - Color that you are experiencing was discussed recently in the link below. Doronhi said in that discussion that as far as he knew the issue emerged in librealsense 2.43.0 and seemed to be based on an issue in librealsense that was not currently fixed. |
|
Hi, |
Thanks for the reply, @doronhi. Yes, sure. I required this following topics, /camera/depth/camera_info - All camera information.
I am getting low FPS, Can you please help me with how to reduce resolution and what is a suitable resolution for subscribing to all this above topic with appropriate FPS and speed. Thanks! |
|
The list of topics you added let me think that an improvement in performance can be made - the colorizer filter is currently applied to both depth and aligned-depth-to-color images, although you use only one of them. This, however, requires some coding, possibly messy and hacky looking... The README.md file describes the parameters used for setting stream resolutions. For example The FPS is really machine-dependent. On my machine, I get a little more the 28fps with your configuration so you'll have to try and see if you can find a balance between fps and resolution that is good enough for you. |
Thanks! for replying and yes, I am using aligned depth only.
28 FPS is good! Well, I am using the below line to obtain topics.
I am obtaining distance from an object through a pinhole camera and converting 2d points to 3d point in ROS. For that, I need an organized PC that. I also have tried to obtain distance with depth channel but it was not that good and accurate as compare to PC. https://answers.ros.org/question/372663/convert-pixel-coordinates-uv-to-pointcloud2-x-y-z-python/#372672 Even with a normal aligned stream, I can see RVIZ is hanging. I think it's an issue with the align_depth flag whenever I am setting it to true then it decreases FPS. I have a good system: https://rog.asus.com/laptops/rog-zephyrus/rog-zephyrus-g15-series/ |
|
Hi, I am having this exact same issue after upgrading to realsense-ros 2.3.0 from 2.2.20 . In the previous version I was able to use the I have traced the breaking change to this commit, which appears to change which depth image the pointcloud is obtained from. |
|
Since version 2.2.24, setting align_depth:=true also creates the pointcloud aligned to the color image. That means, it will have a point for every pixel in the color image, instead of the default behavior of having a point for every pixel in the depth image.
The said modification did not assume you would need a pointcloud aligned to the depth image and aligned_depth_to_color image instead of the original depth image. Is this the case for you? |
|
I'm actually using the L515. I reduced the color resolution to 960x540 instead of 1280x720 and reduced the depth resolution from 1024x768 to 640x480 and that is visualising at 30fps in RViz. For my application I don't need the pointcloud and depth image in the same frame, just the depth and color images in a common frame. As the color image is almost always higher resolution than the depth image, how are depth values for the extra color pixels calculated? |
|
My issue is solved and I am getting 20-25 FPS. When I am subscribing topic in RVIZ at that time it is making my RVIZ slow. What I am doing I am just processing it internally through code and Just focusing on output. For test only I am using rviz. |
|
Can I close this case? |
|
As you are the creator of the case @ranjitkathiriya and there are no other matters linked to this case, it is fine to close it if you are satisifed with the outcome. Thanks! |
Hello there,
I have just now installed the latest version of realsense_ros and I have tried various version which was greater than 2.2.22. When I visualize in rviz I see my rviz is hanging and fps is getting around 2 or 0.
Can you help me to resolve this issue? What may be the issue over here? It was working perfectly before with the older version.
The text was updated successfully, but these errors were encountered: