• State
  • Locked Locked
  • Replies 1 reply
  • Subscribers 100 subscribers
  • Views 340 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: OpenVX support for c7x

one and zero
Mastermind 18256 points

Part Number: TDA4VM

Hi Champs,

with regards to c7x support in our OpenVX implementation:

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/latest/exports/docs/tiovx/docs/user_guide/SUPPORTED_KERNELS.html

The Supported Kernel list in the link above states that currently most kernels are only available for the C66x cores and not the C7x ones. For the C7x the deep learning core is the only ones listed as available. Is this list outdated or is there any plan to provide full support for the C7x in the future? For image preprocessing these kernels would be important.

Thanks and kind regards,

one and zero

  • 0
    TI__Genius 14100 points

    Can you get more details on which specific kernels? Many of these can be represented in form of deep learning layers -

      like filters (sobel, Gaussian) - can be mapped to convolution layers

      add, subtract, product etc can be mapped to Eltwise

      Non max suppression - can be mapped to max pooling

      box filters like average filter can be mapped to either convolution or average pool

      color conversion like YUV to RGB is also supported in TIDL

      mean subtract, scale kind of operations can be mapped to batch norm

      resize can be mapped to resize operator in DL offering

     Some of the kernels which are more for feature formation like edge detection, corner detection, optical flow etc these will be difficult to map to deep learning operators.

    So would like to understand what kind of image pre processing kernels are of your interest

     

    Thanks,

    With Regards,

    Pramod