• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 100 subscribers
  • Views 147 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.

J784S4XEVM: J784S4XEVM: for object detection

Arpit Agarkar
Prodigy 20 points
Part Number: J784S4XEVM


Tool/software:

Hi,

I have J784S4XEVM Processor and would like to use it for object detection. Could you assist me with the detailed steps required to achieve this?

Thank You in advance!

  • 0
    TI__Genius 15780 points

    Hi Arpit,

    We have an intern that addressed this exact issue.  I will move this thread to Rohit and he can give you the details on his project.

    Regards,

    Chris

  • 0
    TI__Prodigy 210 points

    Hello Arpit,

    TI Provides a variety of tools to allow 2d vision tasks including your object detection.

    edgeai-tidl-tools: github.com/.../

    If you take a look at our edgeai-tidl-tools repository and read some of the documentation, you will find a multitude of models meant for detection tasks.

    The process is as follows:

    - Select your model from our model zoo.

    - Compile your artifacts through one of the runtimes OSRT/TIDLRT. We recommend using our open source runtime (in README.md of the repository).

    - Copy your compile intermediate representation of the model to your J7 Processor (binaries)

    - Run inference on your J7 Processor

    These are the core steps you should follow. If you need any help along the way please feel free to reach out.

    Best,

    Rohit

  • 0
    TI__Prodigy 210 points

    Hello Arpit,

    We have recently developed a method of compiling the artifacts necessary for running on your EVM more efficiently.

    You can use this tool for compiling and inferencing from some basic models.

    Please unzip the file with (unzip 11_00_08_00.zip). You can view /11_00_08_00/edgeai-tidl-tools/onnx-model-tool/README.md and follow the instructions!

     

    If you get an error when running the python infer.py…. This may be because there was an issue downloading the models. You should wget the models/prototxts from the urls given in the yaml from example_yamls directory in the onnx-model-tool . And move them to models/public.

     

    This should resolve any issues.

     

    Please use the tool given here:

    tidrive.ext.ti.com/.../8fab69d3-0359-4f41-a7b8-389e9072fad6

    Access Code: 4pt3fuv)

    Best,

    Rohit