YOLO Environment Setup

This chapter applies to manual setup with official images. If you are using the YAHBOOM version of the image, you can ignore this tutorial.

JestonNano

The JetsonNano series requires using the Docker image officially provided by ultralytics due to system version reasons. The Yahboom version of the factory image already has the docker environment pre-installed, so no manual download is needed. This section is only for those who want to manually download and install. Before starting, you need to configure the docker proxy environment to avoid network blocking

Note: On JetsonNano, it cannot be converted to engine format because the current JetPack 4.6.3 / TensorRT 8.2.0.6 cannot directly parse Mod operators

One-click installation using commands

t=ultralytics/ultralytics:latest-jetson-jetpack4
sudo docker pull $t && sudo docker run -it --ipc=host --runtime=nvidia --gpus all ultralytics/ultralytics:latest-jetson-jetpack4

Use the CLI prediction method to test

(Execute in the root@jetson-yahboom:/ultralytics directory)

​x
# Run inference with the exported model
yolo predict model=yolo26n.pt source='https://ultralytics.com/images/bus.jpg'
​
# If you cannot access external networks, you can test with local images
yolo predict model=yolo26n.pt source='/root/yolo26_data/bus.jpg'

We can copy the generated images to the ~/yolo26_data/ directory using the cp command, so they can be viewed on the host machine

JetsonOrin Series

If you are using the Yahboom official image, please skip this step. yolo26 is already pre-installed in the factory image

1. Update the package list, install pip and upgrade to the latest version

   xxxxxxxxxx
   sudo apt update
   sudo apt install python3-pip -y
   pip install -U pip
   

2. Install (update) the ultralytics pip package with optional dependencies

   xxxxxxxxxx
   pip install ultralytics[export]
   

3. Restart the device

   xxxxxxxxxx
   sudo reboot
   

4. Download the yolo26n.pt model

   Ultralytics YOLO26 - Ultralytics YOLO Documentation

   

5. Convert to TensorRT format (accelerate inference)

   After downloading, place it in the /home/jetson/ultralytics/ path

   Execute yolo export model=yolo26n.pt format=engine in the terminal # creates 'yolo26n.engine'