• State TI Thinks Resolved
  • Locked Locked
  • Replies 1 reply
  • Answers 1 answer
  • Subscribers 97 subscribers
  • Views 484 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: Pre-trained Yolo conversion

Indrajith TS
Prodigy 20 points

Part Number: SK-TDA4VM

From my understanding of how the edge ai toolkit converts yolov5 the network was already modified to ti version of yv5 by removing and replacing some of the layers that is not of advantageous to run on edge device and then compilation is done. Suppose we are using a custom trained dl network does the toolkit takes care of the unsupported layers? Also how to do the coversions using edgeai-benchmark tool on yolov8 models.

  • 0
    TI__Mastermind 24041 points

    The Model Compiler tool is available part of tidl understands about unsupported layers at time of model compilation.

    The non supported layers are offloaded to the ARM core at the time of inference and supported to DSP side accelerated inference.

    You can look into edgeai tidl tools github repos for understanding model compilation flow : https://github.com/TexasInstruments/edgeai-tidl-tools

    Please check for Python OSRT examples for getting started.

    Coming to the yolov8 model integration we have not tested this model in our current flow.

    One can follow above listed github repos for custom model compilation purpose.