Usability of Deepstream with Python and Jetson Nano

Hi,

I have a question about the usability of Deepstream working with Python (not C++) on the Jetson Nano:
(I would like to use Deepstream for person detection and counting on the Jetson Nano)

First some information about my system:

  • Jetson Nano
  • image: jetson-nano-4gb-jp441-sd-card-image
  • Ubuntu 18.04
  • Processor ARMv8
  • OS type: 64-bit
  • Graphics: NVIDIA Tegra X1
  • CUDA: 10.2
  • Python 3.6.9

As far as it is possible I would like to avoid using Jupyter Notebook and Docker.
If it is necessary to use Jupyter Notebook or Docker, I would do it naturally!

Question:

  1. Subject: Python Sample Apps Source Details
    https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_Python_Sample_Apps.html#

The following is written in section MetaData Access:
“The bindings are provided in a compiled module, available for x86_64 and Jetson platforms.”
(This is also written here: https://github.com/NVIDIA-AI-IOT/deepstream_python_apps/blob/master/HOWTO.md#metadata_access)

Does this mean that I can’t use this module on the Nano since the Nano has an ARM64 architecture?
Would it follow that I cannot apply this samples (https://github.com/NVIDIA-AI-IOT/deepstream_python_apps) on the Jetson Nano?
It seems to me that all (?) examples somehow access / use the the MetaData library (this would certainly also be necessary for custom applications I think).

Please tell me that I’m wrong.

I have already read a lot of posts within the NVIDIA forum that Deepstream can basically be used on the Nano but the question described above still unsettles me.

Best regards
chris

Hi,
You can install DeepStream SDK:

$ sudo apt install deepstream-5.0

and download the python samples.


Should be runnable without any problem.
1 Like

Hi,

thanks for your reply.

I tried deepstream-test1 and deepstream-test2 yesterday yesterday and both was runable :)

Best regards