Host-container HW and SW compatibilities on Jetson Xavier NX with custom carrier board

Robotics & Edge Computing Jetson & Embedded Systems Jetson Xavier NX
, ,
1

Dear Nvidia community,

we are running an application on edge devices based on Jetson Xavier NX with a custom carrier board. The Jetpack version is 4.4, and we have L4T 32.4.3.
For the future, we would like to run our application in a (docker) container. We are thinking about using the base image nvcr.io/nvidia/l4t-jetpack:r35.4.1, and installing additional dependencies like deepstream and our own libraries on top of it. Another option would be using a deepstream base image, but from what I saw it seems to be lacking some other dependencies.

For some strategic decisions it would be really important for us to answer the following questions:

  1. Is it possible to have a docker container with Jetpack 5.1, while the host system has 4.4? Are there any limitations in this case?
  2. Is it possible to use HW components like GPU (Cuda and Tensor cores), NVDLAs, NV Encoder/Decoder, NV JPEG, VI (ISP and VIC), SPE?
  3. Is it possible to use external HW and interfaces, like Cameras (CSI and USB), serial interface, I2C bus from within the container?
  4. Is there an easy way to verify that a certain HW component is accessible with its full capabilities from within the container?

In previous online research it turned out quite difficult to get information about these questions.
We would greatly appreciate any advice regarding the feasibility of this undertaking.

Best regards
Josef Kaufmann

3

Hi,

1. No, the L4T version need to be aligned.
2. Yes.
3. Yes
4. Below are some containers that can run on Jetson. You can give it a try simply.

Thanks.

1 Like
5

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.