@Jetson YOLO: Ubuntu Desktop and JetPack

· 2 min read

This chapter is about verifying the machine before doing anything more ambitious. On Jetson, the exact JetPack and library stack matters enough that version drift can waste hours later.

Linux Command Line

If you are still new to Ubuntu CLI usage, start with Basic Commands. You do not need to be a shell expert, but you do need to be comfortable moving around the filesystem and installing packages from the terminal.

Check the Desktop Environment

If you see the NVIDIA-styled Jetson desktop, that is a quick sign that the system is at least booting into the expected Linux environment.

Jetson Linux desktop on Jetson Orin Nano

Install jtop

One of the first tools worth installing is jtop from jetson-stats. It gives you a much better operational view of the Jetson module, including clocks, memory, thermal state, and power mode.

sudo apt install python3-pip
sudo -H pip install -U jetson-stats

If you are in mainland China:

sudo -H pip install -U jetson-stats -i https://pypi.tuna.tsinghua.edu.cn/simple

Then reboot:

reboot

After reboot:

jtop

If installation succeeded, you should see a terminal UI similar to this:

jtop dashboard showing CPU, GPU, memory, and system status on Jetson

Confirm JetPack and Library Versions

The next check is more important than it looks. You want to confirm that the system version, JetPack version, CUDA, cuDNN, TensorRT, and related libraries match the environment you plan to target.

jtop info page showing JetPack, CUDA, TensorRT, and system versions

If your environment differs too much from the expected stack, you may need to reflash the device using NVIDIA SDK Manager from a supported x86 Ubuntu host.

In the original setup behind this guide, JetPack 5.1.3 was chosen deliberately because:

  • JetPack 6 was still rough for some dependencies
  • many third-party setup notes and prebuilt wheels still targeted JetPack 5
  • TensorRT and CUDA compatibility was easier to reason about

If you do reflash, use the official SDK Manager workflow:

NVIDIA SDK Manager

This may feel like infrastructure work rather than model work, but on Jetson it is one of the highest-leverage setup steps you can take.