• State Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 3 answers
  • Subscribers 99 subscribers
  • Views 635 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.

TDA4VM: PSDKRA/TIOVX - create user data object using PyTIOVX

sagarkirtany
Genius 4950 points
Part Number: TDA4VM

Hello,

I am working with PSDKRA 6.1.x.x.

I have generated a TIOVX kernel whose input and output are user data objects.

I was able to generate the kernel successfully using the PyTIOVX API.

I also want to generate a use-case which uses this kernel.

But I am not able to find the relevant PyTIOVX API to generate user data objects in the use-case.

Please advise.

Thank you.

  • 0
    Genius 4950 points

    Hello,

    Any updates for this thread?

    Thank you.

  • 0 in reply to sagarkirtany
    TI__Genius 10355 points

    Hi Sagar,

    Sorry for the delay, the PyTIOVX tool can be used to create custom OpenVx kernels but the user will have plug them in a use case manually as of today.

    Although we are working on a GraphBuilder which is planned to be available only after Jan 2021.

    Is there anything specific you are looking for? There are plenty of examples in the vision_apps/apps/basic_demos and dl_demos folder. Hope it serves you as a reference example on how to use it.

    Regards,
    Shyam

  • 0 in reply to Shyam Jagannathan
    Genius 4950 points

    Hello Shyam,

    The TIOVX documentation has a section that describes how PyTIOVX can be used to generate a use-case which in turn instantiates a user kernel. I was able to generate a use-case with these instructions. However, at that time, my kernel was set to accept an input image and produce some scalar outputs. The relevant snippet of the PyTIOVX use-case generation script is shown below.

        width = 640
    height = 480
    input_image = Image(width, height, DfImage.U8, name="input_image")
    out1 = Scalar(Type.INT16, 0, name="output1")
    out2 = Scalar(Type.INT16, 0, name="output2")
    out3 = Scalar(Type.INT16, 0, name="output3")
    graph.add ( NodeMyKernel(input_image, out1, out2, out3, target=Target.DSP1) )

    I modified the I/O parameters of my kernel to accept and produce user data objects. The kernel code was re-generated after these modifications. But I was not able to modify the PyTIOVX script to re-generate the use-case, because the documentation does not list any API to generate user data objects in the graph.

    So is this capability supported through PyTIOVX?

    Thank you.

  • 0 in reply to sagarkirtany
    TI__Genius 13845 points

    Hello Sagar,

    The PyTIOVX tool for application development is experimental in this release.  Therefore, support for the user data objects has not been added.  The source is available in case you would like to modify this to support your use case.

    In future releases of the SDK, there will be tooling which supports generating applications.

    Regards,

    Lucas

  • 0 in reply to Lucas Weaver
    Genius 4950 points

    Hello Lucas,

    I understand what you are saying.

    But if it is possible, please tell me how to modify the source to support my use-case.

    Thank you.

  • 0 in reply to sagarkirtany
    TI__Genius 13845 points

    Hello Sagar,

    Please download the PSDKRA 6.2 release as this has the user data object Python files inside the PyTIOVX module.  This was not included in the 6.1 release but is present in the 6.2 release.

    Regards,

    Lucas

  • 0 in reply to Lucas Weaver
    Genius 4950 points

    Hello Lucas,

    I have working code with PSDKRA 6.1 and do not wish to migrate to PSDKRA 6.2 just yet.

    Is it possible to use the new Python files from PSDKRA 6.2 in PSDKRA 6.1?

    Thank you.

  • 0 in reply to sagarkirtany
    TI__Genius 13845 points

    Hi Sagar,

    Yes, you can use the Python files from 6.2.  I would recommend downloading the 6.2 release and doing a diff of the PyTIOVX folders and pulling in the relevant changes to your 6.1 download.  There are only minor changes in PyTIOVX from 6.1 to 6.2.

    Regards,

    Lucas

  • 0 in reply to Lucas Weaver
    Genius 4950 points

    Hello Lucas,

    I have pulled in the changes as per your instructions.

    I have some other work lined up, so I cannot report immediate progress on this thread.

    So I will close this thread now, and I will open another one if I run into any issues later.

    In the meantime, thank you for your support.