Setting up Hello AI World Containers for Jetpack 5.1.3

[eclark.pgh]

Hello, I’m having some trouble running the correct container for the hello world tutorial, as the current documentation only provides a guide up to 5.1.1. Could I get some guidance on how to find the correct container and run it for this version of Jetpack-- should I clone a different repository or pull something directly through Docker? Apologies for the likely frequent issue regarding the firmware versions, but any help would be much appreciated.
Thanks,
Elliot

[AastaLLL]

Hi,

Please use r35.4.1 for JetPack 5.1.3 since it is a minor release from JetePack 5.1.2.

https://hub.docker.com/r/dustynv/jetson-inference/tags

Thanks.

[eclark.pgh]

I believe it automatically ran that container, although I wasn’t sure it was working correctly since I didn’t see the same UI for choosing which models to download like in Dusty’s tutorial. Thank you for the confirmation.

I am having a problem seeing my camera output however. When I attempt docker/run.sh, I get the output “[OpenGL] failed to open X11 server connection.” and although the camera is connected I can’t see the live video feed. I saw that other users with this problem edited some of the python files directly to resolve this but I believe they were building from source. What would be the best option for dealing with this? Thanks very much

edit: I got it fixed. Had to run the command on the jetson connected to a monitor rather than SSH.

[dusty_nv]

Hi @eclark.pgh , glad you got it working! Yes, the GL/GLX window won’t work over SSH with X11 forwarding, because it uses CUDA interoperability - sorry about that (even if it did, it would be very slow). For remote viewing I recommend using WebRTC, RTP, or RTSP. Also that model downloader tool had been removed in lieu of an on-demand downloader which automatically pulls the model the first time it is loaded.