2-3 Installing the Jetson Component Environment
Learning Objectives
This section helps readers quickly master the installation and configuration process of NVIDIA JetPack 6.2, ensuring the Jetson platform has a complete AI development environment. After completing this section, readers will be able to:
1. Understand the components and purposes of the JetPack SDK, including key modules such as CUDA, cuDNN, and TensorRT.
2. Follow the complete steps to install JetPack and its dependencies on a Pandora system.
3. Verify the installation results and configure a Python development environment.
4. Diagnose and resolve common installation issues (e.g., version compatibility, insufficient disk space).
5. Build a stable software foundation for subsequent AI development (e.g., machine vision, edge inference).
This document provides practical, hands-on guidance suitable for both beginners and advanced developers to quickly set up a Jetson development environment.
JetPack Installation Guide (Including CUDA, cuDNN, TensorRT, etc.)
NVIDIA JetPack provides a complete Software Development Kit (SDK) that integrates core components such as CUDA, cuDNN, and TensorRT, enabling AI, deep learning, and machine learning applications.
Installation Requirements
1. System baseline: Pandora has been flashed successfully (JetPack 6.2 or a compatible version).
2. Network: Connected via Wi-Fi or Ethernet with a stable internet connection.
3. Disk space: At least 20GB free on the system disk (NVMe SSD). A 128GB or larger SSD is recommended.
4. Target platform: Pandora with Jetson Orin Nano (4GB/8GB) or Orin NX (8GB/16GB).
Installation Steps
Step 1: Confirm Network and Update Package Index
1. Ensure Pandora is connected to the internet (Wi-Fi or Ethernet).
2. Open a terminal and run:
Step 2: Install JetPack Packages (CUDA / cuDNN / TensorRT Included)
JetPack is provided as a meta-package that bundles the core development tools. Install everything with:
sudo apt-get install nvidia-jetpack -y
This typically includes:
1. CUDA Toolkit
2. cuDNN
3. TensorRT
4. OpenCV
5. Multimedia API, etc.
Estimated installation time: 10–30 minutes (depends on network speed and storage performance)
Step 3: Verify Installation Status
1. Verify CUDA Toolkit packages:
dpkg -l | grep cuda
2. Verify cuDNN packages:
dpkg -l | grep libcudnn
3. Verify TensorRT packages:
dpkg -l | grep tensorrt
4. Verify OpenCV in Python:
python3 -c"import cv2; print(cv2.__version__)"
If any component does not show up correctly, check your network connection and rerun Step 2.
Step 4 (Optional): Install v4l-utils (V4L2 Camera Utilities)
v4l-utils provides common tools for Linux V4L2 (Video4Linux2) video/camera devices. It is useful for:
- Listing detected camera devices (e.g., /dev/video0)
- Checking supported formats/resolutions/FPS (MJPEG/YUYV, etc.)
- Quick troubleshooting when a camera has no image or format mismatch
Install:
sudo apt-get install -y v4l-utils
Quick checks:
v4l2-ctl --list-devices
v4l2-ctl -d /dev/video0 --all
v4l2-ctl -d /dev/video0 --list-formats-ext
⚠️ Notes
1. Disk space: Make sure at least 20GB is free before installation. A 128GB+ NVMe SSD is recommended.
2. Network stability: JetPack downloads a large number of packages—use a stable connection to avoid interruptions.
3. Installation duration: nvidia-jetpack installation may take 10–30 minutes depending on network and storage speed.
4. Version compatibility: Refer to NVIDIA official Jetson documentation and JetPack compatibility information: JetPack SDK Compatibility