• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 102 subscribers
  • Views 219 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

SK-TDA4VM: In Model Comopser newly created own model does not start when launched from the AI gallery window

Igor Kovryga
Prodigy 190 points
Part Number: SK-TDA4VM
Other Parts Discussed in Thread: TDA4VM

Hello E2E,

Using Model Composer I created my own model. But at the final step - Deploy, the download to my SK-TDA4VM board was unsuccessful, the download process lasted more than 3 hours and was not completed. TI support service advised me to first download the model to my PC and then write it into SK-TDA4VM from my PC. I did this by writing the model into the folder where the SDK models are located /opt/model_zoo. My model has appeared in the Edge AI gallery list, I can select it, but after pressing the Start button nothing happens. My model does not start, but other SDK models from this directory work. I checked my model files and the SDK model files - they are similar, only my model file is larger in size. And my Run.log cosist more additioanal information. Probably the newly created model needs to be initialized? But there is nothing about this in the instructions.
Please help me resolve the issue.

Best Regards

Igor Kovryga

  • 0
    Prodigy 190 points

    Hello E2E,

    I want to add one more thing. I think the problem where the created model is not written to the board directly is due to the browser. The browser itself does not allow you to do this. I tried everything available: Edge, Chrome, Mozilla - The result is always the same. Screen below

  • 0 in reply to Igor Kovryga
    TI__Mastermind 24041 points

    Hi Igor,

    I have assigned this thread to expert, he will get back to you on this.

  • 0
    TI__Intellectual 1710 points

    Hello Igor,

    thank you for your patience.  I tried to reproduce the issue - I have trained a object detection model (yolox_s_lite), compiled it with default settings , and deployed it to the TDA4VM EVM.   Deployment takes around 10s for me.   I also was able to perform a live preview with the trained model.

    Can you provide some additional information:

    • are you training an image classification or object detection model?  which model?
    • what compile settings are you choosing?
    • what SDK version image is your TDA4VM running?   9.0.0? 9.0.1?

    Thanks in advance.

    Martin