Onnxruntime or alternative

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Hey guys,
so i am trying to use face recognition models created in onnx on jetson nano but i have some problems and it would be great if someone could help me :)

My current code (Python) is working on my laptop but i am using onnxruntime there but i cannot get how to install it on jetson nano and what performance compared to old laptop i can expect.
So could anyone point me to one of two things:

  1. some tutorial or code example how to go from onnx model to what i should use on jetson nano - Python3 please :)

  2. How to install onnxruntime on jetson nano?

2

Hi,

1. We always recommend user to convert their model into TensorRT engine for performance.
Please check this sample for the details:
/usr/src/tensorrt/samples/python/yolov3_onnx/

2. onnxruntime should work on Jetson Nano. You can check this topic for more information:
https://devtalk.nvidia.com/default/topic/1050569/jetson-nano/how-to-use-onnxruntime-for-jetson-nano-wirh-cuda-tensorrt-/

Thanks.

3

  1. You provided me with an example for python2 and it’s not even supported anymore by their creators so … maybe you can help me out with something for python current version like python 3.6???

  2. So in the topic you provided is says that a guy has some problem with using it but i don’t understand how to install it. I have to build it from source right?

4

Hi,

Suppose you can build onnxruntime from source with the steps here:
https://github.com/Microsoft/onnxruntime/blob/master/BUILD.md

And apply the Nano GPU support which is mentioned here:
https://devtalk.nvidia.com/default/topic/1050569/jetson-nano/how-to-use-onnxruntime-for-jetson-nano-wirh-cuda-tensorrt-/post/5332020/#5332020

Thanks.