• DepthAICommunity

  • OAK D Lite on Raspberry PI 3 / Ubuntu 16 : worth trying or no go ?

Hi all !

Taking the opportunity of this first discussion to introduce myself : i'm a hobbyist roboticist with a professional IT background, only not in robotics 🙂
My main project is a raspberry pi powered mobile platform which is intended as a test bed for future experiments in social robotics, man-machine interface, experiments in machine learning, and basically having fun with a child dream coming true !

I've received my OAK D Lite last year and i've successfully tested in on several Windows PC.

Now i'd like to implement it in my robot platform which, is powered by a Raspberry pi 3 running ROS 1 / Ubuntu 16

For several reasons i'm keeping this setup as i can't upgrade to Raspberry pi 4 or Ubuntu 18 / 20

Following the installation instructions for ubuntu i saw that the qml-module-qt-labs-platform dependency is missing in the Ubuntu 16 / ARMHF repositories which makes DepthAI install impossible.

In my understansing, Depth AI provides an intergrated user interface to interact with the Oak-D, so i was thinking that it may be possible to access the Oak-D even if Depth-AI is not installed on my setup.
Is there a way i could use direct USB communication with the Oak-D in Python ?

Is it worth looking into it ? Or shall i rather consider using the Oak-D in different projects on more recent platforms ?

Thank you for your input

  • erik replied to this.

    Hi Caz ,
    Sounds like a cool project!

    In my understansing, Depth AI provides an intergrated user interface to interact with the Oak-D, so i was thinking that it may be possible to access the Oak-D even if Depth-AI is not installed on my setup.

    Yep that's correct. You can use just depthai library (not the demo) which doesn't require QT, and we have a bunch of depthai python examples here. depthai library will communicate with the OAK-D through USB communication (through XLink protocol).
    Thoughts?
    Thanks, Erik

      Thank you Erik !
      I've checked the link, i see it's for Python 3.
      ROS 1 default is Python 2, so i think i'll take a two steps approach : first, test on an up-to-date non-ROS system with Python 3; then attempt to have my ROS run Python 3, i've found a page that seems encouraging with a step by step guide.

      I'll check all this next week-end. Thank you for your help !

        Hi Caz , Sounds great, let us know if you have any additional issues/questions🙂!

          I've just tried upgrading my setup to run python 3; my conclusion is that the oak-d is a no go on ROS 1 / Raspberry pi3.

          I'd love it if someone ould prove me wrong; but here's how i got there.
          As Python 3 is necessary i found this tutorial explaining how to configure ROS 1 to run Python 3 instead of default Python 2

          (apologies if there's a no-link policy i missed, can be removed in this case!)
          But the page is about ROS Melodic, which is based on Ubuntu 18.

          When trying to follow the indicated steps nontheless i met several pip syntax errors which when googled point to compatibility issues that could involve the Python 3.5 version i use
          Python 3.6 is not available anymore for my environment in any directory i've tried

          Moving to Ubuntu 18 is not an option on my Pi 3 (already tried, so slow it's not usuable), so for me it's a no go in my current state.

          This may rush switching to Raspberry pi 4 when/if they get back to regular prices so i can use the Oak-D in my project !

          • erik replied to this.

            Hi Caz
            What about using depthai C++ instead of python? Then should solve the issues. As all of depthai ros examples/drivers are written in c++ as well. Thoughts?
            Thanks, Erik

              That's an interesting option, i'll look into it !

              It's been decades since i haven't compiled / built C++ though 🙂 It might be the motivation i'd need to get back into it!

              But ROS adds complexity, the OS and env versions i use don't help, so there may still be dependencies issues i'd have troubles solving.

              I'll have a look at it this week-end, any help pointing me in the right direction would be greatly appreciated !

              I'll also try installing ROS 2 on a pi 4 i have and installing depthAI on it.

              • erik replied to this.

                Hi Caz ,
                Probably using pi4/ros2 will be the easiest way to resolve these version issues 🙂
                Thanks, Erik

                  24 days later

                  Hi again !
                  I took the time to setup a brand new environment, RPI 4 2GB, Ubuntu 20.04, ROS 2

                  I've installed DepthAI prerequisites and i'm now attempting to run the Installing Depth AI script from the page
                  https://docs.luxonis.com/en/latest/pages/tutorials/first_steps/
                  where i pick the Linuc tab and execute the indicated script.

                  It goes well up to one point where it gets stuck.
                  All that goes well :

                  Collecting pyqt5
                  Using cached PyQt5-5.15.9.tar.gz (3.2 MB)
                  Installing Build dependencies ... done
                  Getting requirements to build wheel ... done

                  and it gets stuck on the next line ;
                  Preparing metadata (pyproject.toml) ... -

                  Won't do a thing from there, apparently uses all CPU after a while and won't get past this.

                  First noob question that comes to mind : should the Oak-D be connected to the Raspberry at that point of the install ?
                  I've not tested that yet 🙂

                  Thank you for the support !

                  • erik replied to this.

                    Hi Caz ,
                    You don't need pyqt5 per se (you only need it for depthai-demo), that's why by default we don't use it for depthai-demo, and use OpenCV "gui". So you can skip this requirement. In general, I'd suggest using preconfigured RPI image, especially if you aren't exactly sure what you are doing.

                    should the Oak-D be connected to the Raspberry

                    It doesn't really matter.
                    Thanks, Erik

                      Thank you Erik,

                      i was worried that all required dependencies would not install properly if the script didn't go through all the way as there might have been other steps skipped if it stuck there;.
                      So i went on installing for Python; i just tested, it works great and i can run examples !

                      I considered using a preconfigured image as a last resort, but i was aiming at an Ubuntu 20 / ROS 2 environment.
                      It is now complete with DepthAI on top !

                      Thanks again 🙂

                        Hi Caz ,
                        Great to hear you got it working🙂!
                        Thanks, Erik

                          a year later

                          Caz Hello! I am running into issues myself when trying to follow the getting started guide. I am using Ubuntu 22.04 so things might be different for me. Do you remember what all you did to get it up and running?

                          I was able to get it working to an extent at one point. I could only get the preview running for 10-20 seconds before crashing. Every other program crashed it. I’ve tried raspbian and Ubuntu and I just can’t seem to get this thing working.

                          • Caz replied to this.

                              Hi @Kato
                              Please be more specific as to what errors you are receiving then the program crashes. This could give us a little more insight into what might be missing.

                              Thanks,
                              Jaka

                              Kato Hi,

                              it's been one year already, since then i've had to switch to 3D printing and hardware issues and i didn't get back to the PI 4 environment yet.

                              Reading this thread again, i can't remember exactly what i've had to do to get it going, but i remember that "Install for Python" was a specific install procedure that i must have found online and followed.
                              Unlike the previous installs i tried, that install worked fine for me so i guess i didn't have to tweak anything.

                              But as jakaskerl mentionned, it would certainly be helpful if you could provide more insights on the issues you encounter.