Object detection using verdin & python

Technical Support
1

Subject:Verdin Development Board (Verdin iMX8M Plus) Queries

Dear all,

I am currently working with the Verdin Development Board (Verdin iMX8M Plus) and would appreciate your guidance on the following queries:

  1. Using the Toradex Easy Installer, can we install a standard Linux OS directly (not Toradex Linux or TorizonCore)?
  2. Is it possible to run Visual Studio Code directly on the Verdin board, similar to a full Linux OS with GUI?
  3. How can I capture an image from a USB camera using Python? (I intend to use the NeoAPI Python library.)
  4. How can I process images using deep learning models in Python on the board?
  5. If we develop a GUI using Python (for example, with Tkinter), will it display properly on a connected monitor or display?
  6. We are currently testing with TorizonCore OS and can connect via SSH. How can we upload and update our Python program on the board?
  7. We are trying to connect to the Verdin board using the Torizon Extension in Visual Studio Code, but the device does not appear. How can we resolve this?
  8. Is it possible to update or upload programs offline using an RJ45 Ethernet cable (PC to Verdin board directly)?
  9. Can we customize the OS on the board (e.g., install a full Linux distribution or even Windows)?

Kindly provide your guidance on these points.

2

Hi, I’m the Toradex AI Assistant. The Toradex Support team will follow up with you shortly!

Did you like my answer?

:+1: :-1:

Hello!

For deep learning and object detection on the Verdin iMX8M Plus, here is some guidance based on the available documentation:

  1. Operating Systems: The Verdin iMX8M Plus comes with the Toradex Easy Installer pre-installed. The supported operating systems include:

    • Torizon OS
    • Yocto Project Reference Images

    You can find more details about the hardware and supported operating systems here.

  2. Machine Learning Support: For machine learning applications, Toradex supports several frameworks and tools:

    • MXNET Framework: Suitable for neural network models.
    • GluonCV: Implementation for SSD (Single Shot Detection).
    • AWS SageMaker Neo: Used for cross-compiling models to the i.MX8 platform.

    More information on these tools can be found here.

  3. Model Execution: To execute models on the board:

    • Use the DLR runtime with Torizon.
    • Ensure models are compiled specifically for the nxp_i.MX8QM target.

  4. Development Tools: For development, the following tools are supported:

    • Torizon IDE Extension for VSCode: Facilitates development on Torizon.
    • Torizon OS Customization: The board supports customization of the Torizon OS.

    Additional details can be accessed here.

Regarding your questions about standard Linux distributions, direct VSCode installation, and Windows, these configurations are not covered in the available documentation. However, the resources provided should help you get started with the supported configurations.

If you need more specific details about any of these configurations, feel free to ask!

Best regards,
Toradex AI Assistant.

3

Hi @sandy1464!

I’ll answer your questions:

  1. It isn’t possible, as far as I know.
  2. No, it isn’t possible. You could run VSCode on your host machine and use our Visual Studio IDE Extension to help develop your project.
  3. I don’t have the necessary expertise about USB cameras, but I suggest you take a look at our Partner Network, where you can find a partner who will have the knowledge needed to accelerate your project.
  4. We have this article on our developer site about machine learning on Toradex CoMs, but I would suggest searching for a partner (in the link I sent on the last topic) that can help you with this specific need.
  5. Yes, you’ll need to do some modifications, for example, on the device tree overlays, but it will display your GUI without problems.
  6. You can copy the .py file to the board using the scp command, for example. Besides that, we have the Torizon Cloud, which has a lot of interesting features like OTA updates that can be used to update the Python script.
  7. The device appears on the extension if it is connected to the same network as the host machine. Here, you can find a step-by-step guide on how to connect to the IDE Extension.
  8. As far as I understand, yes. You would need a local network between the host machine and the module, but once it is done, you can use the scp command, for example.
  9. You can create a custom image using Yocto Project, and some of our modules support Windows Embedded Compact, but the Verdin family doesn’t.

I hope I could help you with these answers.
Feel free to ask if you have any doubts.

Best regards.
Lucas Azeituno

4

Thank you.