Official NVIDIA Libraries and SDKs for Jetson Orin Nano
These are the core libraries and modules provided by NVIDIA as part of the JetPack SDK, optimized for the Jetson Orin Nano's hardware (Ampere GPU, ARM CPU). They are pre-installed or easily loaded via SDK Manager or JetPack images. JetPack 6.x (latest as of October 2025) includes support for up to 67 TOPS AI performance on the "Super" variant. Use SDK Manager on an Ubuntu host to install:
NVIDIA SDK Manager.
| Category | Library/Module | Description/Use Case | Link |
|---|
| AI/ML Frameworks | TensorRT | High-performance deep learning inference optimizer; accelerates neural networks for real-time AI apps like object detection. | TensorRT Documentation |
| AI/ML Frameworks | cuDNN | GPU-accelerated library for deep neural networks; essential for training/inference in CNNs, RNNs on Jetson edge devices. | cuDNN Download |
| AI/ML Frameworks | CUDA Toolkit | Parallel computing platform and API; enables GPU-accelerated compute for AI, graphics, and simulations (CUDA 12.4+ on JetPack 6.0). | CUDA Toolkit |
| Computer Vision | OpenCV (NVIDIA-optimized) | Image/video processing library with GPU acceleration; used for feature detection, tracking in robotics/drones. | OpenCV on Jetson (via JetPack) |
| Computer Vision | VisionWorks | Advanced vision pipeline toolkit; accelerates computer vision tasks like stereo matching on Jetson hardware. | VisionWorks |
| Multimedia | Multimedia API (V4L2, GStreamer) | Hardware-accelerated video decode/encode; for streaming, camera pipelines in intelligent cameras. | Multimedia API Docs |
| Robotics | Isaac ROS | Modular robotics framework with accelerated nodes for perception, navigation; builds on ROS 2 for Jetson Orin Nano robots. | Isaac ROS |
| Developer Tools | Nsight Systems/Compute | Profiling and debugging tools; analyzes GPU/CPU performance for optimizing AI apps. | Nsight Tools |
| System/Base | Jetson Linux (L4T BSP) | Linux for Tegra base OS (Ubuntu 22.04); includes kernel, bootloader, drivers for Orin Nano hardware. | Jetson Linux Docs |
| System/Base | JetPack SDK | Full stack including all above; for end-to-end AI app development/deployment. | JetPack SDK |
| GPIO/Peripherals | Jetson.GPIO | Python library for GPIO control; mimics RPi.GPIO for digital I/O in maker projects/sensors. | Jetson.GPIO GitHub |
Installation Notes: Flash via
Jetson Orin Nano Getting Started Guide or
Downloads Center. Supports microSD, NVMe SSD, or eMMC storage.
Free Third-Party Modules and Libraries
These are open-source, community-maintained libraries compatible with Jetson Orin Nano (aarch64 architecture, JetPack 6.x). They extend functionality for AI, robotics, and edge computing. Install via pip, apt, or GitHub builds (use conda for Python deps to avoid conflicts). Many are tested on Orin via NVIDIA forums/GitHub.
| Category | Library/Module | Description/Use Case | Link |
|---|
| AI/ML Frameworks | PyTorch (Jetson Wheels) | Deep learning framework with CUDA support; for model training/inference on edge (use NVIDIA-provided wheels for aarch64). | Jetson PyTorch Wheels |
| AI/ML Frameworks | DeepSpeed | Microsoft library for distributed training/large models; optimizes memory for LLMs on Jetson (install via pip in conda). | DeepSpeed GitHub (build for aarch64) |
| AI/ML Frameworks | MMCV (MMDetection) | OpenMMLab's computer vision lib; for object detection/segmentation pipelines (full version for PyTorch). | MMCV GitHub (Jetson build guide) |
| Computer Vision | PyTorch3D | Facebook's 3D deep learning lib; for mesh rendering/point clouds in AR/VR on Jetson. | PyTorch3D GitHub (conda install) |
| Computer Vision | OpenEXR | Industrial light/magic's HDR image lib; for multi-channel image I/O in VFX/graphics apps. | OpenEXR GitHub (apt or build) |
| Robotics | ROS 2 (Humble/Iron) | Robot Operating System; middleware for robot control/perception (accelerated with Isaac ROS). | ROS 2 on Jetson (via apt) |
| Mesh/Geometry | MeshPlex | Geometry processing lib for meshes; used in 3D modeling/simulation on embedded systems. | MeshPlex GitHub (pip install) |
| Audio/ML | Mido/MIDIUtil | MIDI file handling libs; for music generation/AI audio projects (lightweight for edge). | Mido GitHub (pip) |
| Data Processing | Pandas/NumPy/SciPy | Data manipulation/scientific computing; core for ML data prep (pre-built wheels for Jetson). | Jetson AI Lab Pandas Guide (pip) |
| Ecosystem Projects | Jetson AI Lab Examples | Generative AI demos (e.g., Stable Diffusion); showcases third-party integrations like Hugging Face. | Jetson AI Lab |
| GPIO/Peripherals | Adafruit CircuitPython | Sensor/peripheral libs; for I2C/SPI devices in IoT projects (compatible with Jetson.GPIO). | Adafruit Blinka (pip) |
| Build Guides | Jetson Orin Nano Super Guide | Community guide with third-party optimizations (e.g., TensorRT-LLM builds). | ajeetraina GitHub |
Notes: For installation issues (e.g., PyTorch3D/MeshPlex), use conda environments on JetPack 6.0 (Python 3.10). Check NVIDIA Forums for aarch64 builds:
Jetson Forums. Check compatibility with JetPack version via
Jetson Compatibility Matrix.
