Please tell me how to install RealsenseSDK

[drinkBr]

I'm using Ubuntu 22.04 LTS on Raspberry Pi 4B and I want to install the Realsense SDK, but I've tried several methods on this GitHub repository and none of them seem to work. The realsense-viewer is also very slow and practically unusable. Can you please suggest the best method?

[MartyG-RealSense]

Hi @drinkBr I would first recommend using the libuvc backend installation method with Raspberry Pi if you have not tried it already.

https://github.com/IntelRealSense/librealsense/blob/master/doc/libuvc_installation.md

[drinkBr]

Thank you for your reply.
I completed the libuvc backend installation method with Raspberry Pi. As a result, I can now view both RGB images and depth using realsense-viewer.
Thank you.

[MartyG-RealSense]

You are very welcome. I'm pleased that I could help. Thanks very much for the update!

[drinkBr]

I have one more question. Will this method also work with Ubuntu 18.04 LTS? I tried the same method, but I can only get a 2D depth image. If there are other methods available, please let me know.

[MartyG-RealSense]

Yes, it should also work with Ubuntu 18.04.

Raspberry Pi is a difficult platform to install librealsense on. If the libuvc method works for you then I would recommend editing the libuvc build script to add additional build flags to the CMake build instruction if you need to customize the installation.

[drinkBr]

Thank you for your reply.
I tried the same method, but I can only get a 2D depth image. Can you please suggest the best method?

[MartyG-RealSense]

IF you are using the RealSense Viewer tool then you can switch to 3D pointcloud mode by clicking on the word '3D' in the top corner of the Viewer window.

[drinkBr]

I know that. However, even when I connect the camera, it doesn't show up. In 2D, only the depth image is displayed, and the RGB image shows 'No image'. In 3D, only the axes are displayed.

[MartyG-RealSense]

If the pointcloud is not appearing in 3D mode and only the axes are shown then please try disabling the Viewer's GLSL settings using the instructions at #8110 (comment)

[drinkBr]

Thank you.
When I checked, GLSL was already disabled.

[MartyG-RealSense]

Raspberry Pi can provide depth and RGB streams when used with a RealSense camera but typically has problems when asked to provide more than that, such as depth-color alignment and pointclouds.

Are you able to launch the RealSense SDK's example program rs-pointcloud to test whether a pointcloud can be generated outside of the Viewer?

[drinkBr]

Thank you for your reply!
I'm sorry for the delayed response. When I run it, I get the message 'RealSense error calling rs2_pipeline_wait_for_frames(pipe:~): Frame didn't arrive within 15000'. I read the following link for a solution, but I couldn't understand it. I would appreciate some advice. I'm a beginner when it comes to Ubuntu, realsense, and Raspberry Pi, so I would like detailed guidance as much as possible.
#6518

[MartyG-RealSense]

Frame didn't arrive within 15000 means that the program did not receive new data from the camera for a period of 15 seconds.

The limits of RealSense performance on Pi boards means that an example that does not use pointclouds or alignment might work better for you. Please try running the example rs-capture which displays the depth and color streams individually.

You could also try increasing the size of your Pi's swapfile to create more 'virtual memory' from space on your storage drive that can be used when the real memory capacity on your Pi is consumed. The link below has a swapfile guide for Ubuntu 22.04.

https://www.digitalocean.com/community/tutorials/how-to-add-swap-space-on-ubuntu-22-04

[drinkBr]

Thank you for your reply!
I tried running rs-capture, but I received the following error messages: 'ERROR [281472208197888] (synthetic-stream.cpp:58) Exception was thrown during callback: rscuda9alloc_devI4int2EESt10shared_ptrIT_EiEUlPS1_E_' and 'ERROR [281473096423680] (rs.cpp:237) [rs2_pipeline_wait_for_frames( pipe:~ ) UNKNOWN]'and 'RealSense error calling rs2_pipeline_wait_for_frames(pipe:~): Frame didn't arrive within 15000'. I created a swap of 4GB, same as RAM, and ran rs-capture, but still encountered the same error messages.

[MartyG-RealSense]

Which RealSense camera model are you using, please?

Are you using the official 1 meter long USB cable supplied with the camera?

Does it make a difference if you unplug the micro-sized end of the USB cable from the camera, reverse the connector's orientation and plug it back into the camera (USB-C cables are two-way insertion at the micro-sized end).

[drinkBr]

Thank you for your reply.
I'm using "intel Realsense Depth Camera D435" and the official 1 meter long USB cable supplied with the camera.
I tried flipping the connector and plugging it back into the camera, then ran rs-pointcloud, rs-capture, and realsense-viewer again, but the results didn't change.

[MartyG-RealSense]

Thank you very much.

Next, please try resetting your D435 to its factory-new default settings in the RealSense Viewer using the instructions at #10182 (comment) if you are able to start the depth stream in the Viewer.

[drinkBr]

Thank you for your reply.
I could complete resetting. There was no change, particularly.

[MartyG-RealSense]

Let's try a much simpler test program without any graphics to see whether the problem is that the Pi is being over-burdened by the demands of running graphical applications. Please run the example rs-hello-realsense. This program prints depth as a plain text value without any visuals.

Do you still receive Frame didn't arrive within 15000 when running that simple program?

[drinkBr]

Thank you for your reply.
When I ran rs-hello-realsense, the same error message was displayed.
I'm considering using ROS in the near future to acquire and save image information from the Realsense camera. Do you think the current issue could affect this plan? Also, is there a possibility of improvement through the introduction of realsense-ros? If so, I'd like to give it a try.

[MartyG-RealSense]

RealSense's ROS interface runs as a 'wrapper' on top of the RealSense SDK. So if you have problems with the SDK then it is likely to have an impact on the performance of the ROS wrapper too, unfortunately.

Researching use of Ubuntu 22.04 on Raspberry Pi 4 indicates that other users have experienced very slow performance too, Apparently, if Ubuntu is run off an SD card instead of an external SSD drive then it can be especially slow.

An Ubuntu 22.04 peformance improving guide at the link below suggests that using something called Zswap instead of a swap file can help.

https://teejeetech.com/2022/06/04/tweaks-for-ubuntu-22-04-on-raspberry-pi-4/

[drinkBr]

Thank you for your reply!
It might be because I'm using Ubuntu 18.04, but I couldn't disable 'zswap' and enable 'zram'. I was able to overclock, but I couldn't run a stress test, so I haven't confirmed it.