Hello Toradex Community,
i am currently working on my first project with toradex hardware. I am using the
Toradex Maivin Full Stack Vision System
IMX8M-Plus
I have set it up like this:
- Working on Windows 10
- Connected USB-RS485 converter with A->A and B->B
- Connected ethernet cable from Maivin to USB-Ethernet adapter on my machine with default settings
- Connected Power to GND and VIN (blue light is on)
Now i have the following problems:
- Using “ssh torizon@verdin-imx8mp-xxxxxxx” results in unknown hostname
- Trying to connect with the COM-Port results in random output in the console. I tried using Putty and Tera Term 5
I have investigated the following behaviour:
- Network adapter prompts “Unidentified network”
- Lights of RJ45 connector on Maivin are not beeing turned on (orange and green led)
- Switched A->A to A->B and B->B to B->A on the RS485 (no change in behaviour)
I hope you can understand my problem and may help me out with this.
Best regards and thank you in advance,
Simon
Greetings @Reimonder,
Just to clarify, outside of the issues with SSH and the debug UART, is the Mavin itself actually working, are you able to access the UI?
Using “ssh toradex@verdin-imx8mp-xxxxxxx” results in unknown hostname
The hostname would actually be torizon@verdin-imx8mp-xxxxxxx as documented here: https://support.deepviewml.com/hc/en-us/articles/5236697083277-SSH-into-the-Maivin
Trying to connect with the COM-Port results in random output in the console. I tried using Putty and Tera Term 5
Did you configure/setup the serial connection correctly as documented here: https://support.deepviewml.com/hc/en-us/articles/4429905733005-Debug-Console-UART
Best Regards,
Jeremias
Hey Jeremias,
thank you for your quick response.
are you able to access the UI?
No, im not able to access the UI or ping the Maivin. Blue light of the Maivin is light up.
The hostname would actually be torizon@verdin-imx8mp-xxxxxxx as documented here:
Im using the torizon@, i made a typo when creating the post, sorry.
Did you configure/setup the serial connection correctly as documented here:
I am using the UART_1 on the rear terminal block with the quick wire connector block
I cofigured the COM-Port like in the documentation but i still do not get access to the device.
I opened up the device today to check if everything is connected inside. It looks good and all connectors are connected.
I really don´t know what to try next, because for me i have tried everything what i did with my previous imx8mp on the verdin carrier, it worked for me there.
I hope you could help me out there.
Best regards and thank you in advance,
Simon
Hello @Reimonder ,
According to this page:
Maivin uses DHCP , so you should set up a DHCP server.
Best regards,
Josep
Hi Josep,
Maivin uses DHCP , so you should set up a DHCP server.
Using a DHCP-Server for the Maivin brought no change, no ping possible and no lights on the RJ45 connector.
But i had a small breakthrough with the UART_1. I re-installed the drivers for the adapter and opened the com port and this showed up
The console spammed a for a certain amount of characters after i pressed a on my keyboard and then displayed the login prompt. After i pressed t, because i wanted to enter the username, it spammed to for the same amount of characters.
I cannot login because of the spamming but if i find out anything new, i´ll let you know.
I got connection with SSH now, the problem was, that the hostname at the back of the device was the wrong one. I used the correct one and had immediate access to the device via SSH.
Now i tried starting the demo container like stated in the Introduction [https://support.deepviewml.com/hc/en-us/articles/23493293334157-Native-Segmentation-Application].
sudo systemctl start segmentation
If i use the command to start the segmentation, the device will run into an exception and restart itself, closing the SSH-connection as well with it.
Anyone had this issue before?
Does this service give any useful logs if you try something like journalctl -f -u segmentation after you start it? This service came from Au-zone so we’re not too familiar with it’s inner workings.
Best Regards,
Jeremias
I got it up and running, somehow the camera could not be accessed by the segmentation sample.
I would have another question. Im relatively new to NXP systems but i wanted to ask if there is any How-To guide for setting up an existing pretrained model for the IMX8M-Plus. I have pretrained modely in YOLOv8 and Faster-RCNN which i want to run an inference on the cameras video capture.
My current plan is to install the L6.1.36_2.1.0_MX8MP Firmware on the Maivin and start a Vision Pipeline with the Vision Pipeline Runner of the eIP Toolkit.
Hopefully someone can help me out here.
Thank you very much in advance.
I got it up and running, somehow the camera could not be accessed by the segmentation sample.
That’s great to hear!
I would have another question. Im relatively new to NXP systems but i wanted to ask if there is any How-To guide for setting up an existing pretrained model for the IMX8M-Plus.
Not sure if we have a pre-made how-to depending on what you want to do here. This is our main page regarding AI/computer vision topics: AI, Computer Vision and Machine Learning on Toradex Computer on Modules | Toradex Developer Center
It links to other more in-depth articles as needed. Does any of the content here work for you? Or are you looking for something different?
My current plan is to install the L6.1.36_2.1.0_MX8MP Firmware on the Maivin and start a Vision Pipeline with the Vision Pipeline Runner of the eIP Toolkit.
Just to clarify, you said “eIP” did you mean “eIQ” instead?
Best Regards,
Jeremias
Hi, sorry for my late reply I was off work because of some exams.
Not sure if we have a pre-made how-to depending on what you want to do here. This is our main page regarding AI/computer vision topics: AI, Computer Vision and Machine Learning on Toradex Computer on Modules | Toradex Developer Center
Thats helping a lot, thank you.
Just to clarify, you said “eIP” did you mean “eIQ” instead?
Yes, sorry, that was a typo on my end.
So now if i want to install Torizon core on it, how do i have to set it up hardware wise? Because as far as i know, the installer needs an USB connection to the device to install torizon.
Thanks in advance and best regards,
Simon
So now if i want to install Torizon core on it, how do i have to set it up hardware wise? Because as far as i know, the installer needs an USB connection to the device to install torizon.
It seems Au-zone has documentation on how to load our Easy Installer onto the Maivin: https://support.deepviewml.com/hc/en-us/articles/4417259322893-Maivin-System-Recovery
Best Regards,
Jeremias
It seems Au-zone has documentation on how to load our Easy Installer onto the Maivin: https://support.deepviewml.com/hc/en-us/articles/4417259322893-Maivin-System-Recovery
that worked well, thank you jeremais. But it does not seem like they have any version with pre-installed docker as far as i can see.
But it does not seem like they have any version with pre-installed docker as far as i can see.
I’m not sure I know what you mean here. What kind of OS image are you looking for exactly?
Best Regards,
Jeremias
I’m not sure I know what you mean here. What kind of OS image are you looking for exactly?
I installed the image Torizon for Maivin which was listed in the Eazy Installer and that had no docker package included. But i found another OS image which has docker included. Now i got everything up and running. Thank you for the support.
I see, glad I was able to help!
Another day, another problem.
I got my inference application running with static images using a docker volume. Now i want to access the camera with openCV.
I try to use the camera like this
def generate_frames():
capture = openCV.VideoCapture("dev/video0")
print("capture active: " + str(capture.isOpened()))
while capture.isOpened():
success, frame = capture.read()
if not success:
break
# Capture frame on button press
if openCV.waitKey(1) & 0xFF == ord('e'):
global current_frame
current_frame = frame.copy()
ret, buffer = openCV.imencode('.jpg', frame) # Encode frame
frame_bytes = buffer.tobytes()
yield (b'--frame\r\n'
b'Content-Type: image/jpeg\r\n\r\n' + frame_bytes + b'\r\n')
print("Frame generated, releasing capture")
capture.release()
This image i want to host in the browser using a flask application like this
@app.route('/', methods=['GET', 'POST'])
def index():
return Response(generate_frames(), mimetype='multipart/x-mixed-replace; boundary=frame')
But i am only getting a white screen as a camera image. I tried it on my windows host using the cv2.VideoCapture(0) and it displayed the stream of my webcam.
Are there any things i need to do for the Maivin camera to display it´s image, like installing a seperate driver for it?
Thank you in advance
Hello @Reimonder,
Can you give me more details on which image you installed to the device?
The image which should be used for Maivin is Torizon for Maivin. This image includes the driver for the camera on the device.
This image is available from the Toradex feed as well as from the Au-Zone’s feed (https://maivin.deepviewml.com/image_list.json).
No container is bundled in these images, but Docker is set up and working.
With Torizon for Maivin installed, you can try to run one of Au-Zone’s Demos to confirm everything is working as expected.
Best Regards,
Bruno
Hi @bruno.tx,
thank you for your quick reply. I have installed the TorizonCore 5.7.1-devel-20221026172853+build.0 which was listed in the imagelist you have provided. I will try to run a container of the Au-Zone´s demo apps, thank you for that idea.
I will reply as soon as i have confirmation that the installed os is indeed one working with the demos of Au-Zone.
Thats the exact image i have installed
Best regards,
Simon
I tried to run the Faceblur-sample from Au-Zone but if I start the container i get the following error message:
ERROR: from element /GstPipeline:pipeline0/GstV4l2Src:v4l2src0: Device ‘/dev/video0’ is not a capture device.
I don´t know if i have installed the correct OS on my Maivin but that´s the only one listed as Torizon for Maivin in the Eazy Installer.
Those three images come from the https://maivin.deepviewml.com/image_list.json. As far as i see, this ones do not come with docker installed.
So I installed the one from the http://tezi.toradex.com/image_list_3rdparty.json to use my Maivin in combination with docker.
I don´t know if there is any mayor difference between the Version 5.7.1 and 6.4.0.1. If yes, please let me know
Hello @Reimonder,
I will try to reproduce the issue you faced.
There are major differences between Torizon for Maivin 5.7.1 and 6.4.0.1, but both should work with the Au-Zone demo.
Those three images come from the https://maivin.deepviewml.com/image_list.json. As far as i see, this ones do not come with docker installed.
Do you face any issues when trying to use docker with these images?
Best Regards,
Bruno
