Software to load on a NVIDIA Jetson Nano AI, robotics, and general development.

Flynn · Sep 25, 2025

Software to load on a NVIDIA Jetson Nano AI, robotics, and general development.

1. Core System

JetPack SDK (NVIDIA’s official software stack for Jetson)
- Ubuntu-based OS (usually Ubuntu 18.04 or 20.04 depending on JetPack version).
- CUDA Toolkit (GPU computing).
- cuDNN (deep neural network acceleration).
- TensorRT (inference optimization).
- Multimedia API (camera, video encode/decode).
- OpenCV with CUDA support.

You can flash JetPack with NVIDIA SDK Manager from a host PC.

2. Development Tools

Python 3 (comes by default, but update it).
pip & virtualenv (for Python packages).
CMake & GCC (for compiling code).
git (version control).
VS Code (or nano/vim if you prefer lightweight editors).
JupyterLab / Jupyter Notebook (interactive coding).

3. Machine Learning / AI Libraries

PyTorch (Jetson-optimized wheels available from NVIDIA/JetsonHacks).
TensorFlow Lite (lightweight for edge AI).
ONNX Runtime (for running ONNX models).
scikit-learn (general ML).
pandas / numpy / matplotlib (data science stack).
Hugging Face Transformers (if doing NLP).

4. Robotics / IoT Packages

ROS/ROS2 (Robot Operating System) (if doing robotics).
MQTT (Mosquitto client/server) (for messaging).
Node-RED (visual programming for IoT).
OpenCV (with GStreamer support) (for vision tasks).
PySerial (if using Arduino/serial devices).

5. Cloud / Communication

gRPC (for inter-process communication).
Flask / FastAPI (for REST APIs).
Docker (for containerized deployments).
NATS or RabbitMQ (if you want enterprise-grade event bus).

6. Optional but Useful

Jetson.GPIO (to work with GPIO pins).
DeepStream SDK (for multi-camera / video analytics).
Open3D / PCL (for 3D point cloud processing).
YOLOv5/YOLOv8 repo (ready-to-go object detection).
Edge Impulse CLI (if you want to deploy Edge Impulse models).

Quick Install Order Suggestion:

Flash JetPack (includes CUDA, cuDNN, TensorRT, OpenCV).
Update & upgrade system (sudo apt update && sudo apt upgrade).
Install Python + pip + virtualenv.
Install PyTorch + TensorFlow Lite.
Install OpenCV, scikit-learn, pandas, numpy.
Add extras: ROS, MQTT, Node-RED, Docker, DeepStream (as needed).

Jetson Nano setup script you can run (piece by piece or as a whole). This will give you a clean AI/IoT workstation right on your Nano.

Jetson Nano Install Commands

1. System Update

sudo apt update && sudo apt upgrade -y sudo apt install -y build-essential cmake git curl wget unzip

2. Python & Dev Tools

sudo apt install -y python3-pip python3-venv python3-dev pip3 install --upgrade pip setuptools wheel pip3 install virtualenv jupyterlab notebook

3. Core AI Libraries (Jetson-optimized)

(Pytorch/TensorFlow have special wheels for Jetson)

# PyTorch (use NVIDIA-provided wheel, example for JetPack 4.6 / Nano) wget https://nvidia.box.com/shared/static/xxxxxx.whl -O torch.whl pip3 install torch.whl # TorchVision (compatible version) sudo apt install -y libjpeg-dev zlib1g-dev pip3 install torchvision # TensorFlow Lite pip3 install tflite-runtime # ONNX Runtime pip3 install onnxruntime 

(Replace the xxxxxx.whl link with the current wheel from JetsonHacks/PyTorch repo)

4. Data Science Stack

pip3 install numpy pandas matplotlib seaborn scikit-learn pip3 install opencv-python pillow pip3 install jupyterlab

5. Robotics / IoT

sudo apt install -y mosquitto mosquitto-clients pip3 install paho-mqtt pip3 install pyserial sudo apt install -y ros-<distro>-ros-base # (ROS1) change <distro> to melodic/noetic depending on Ubuntu

6. Cloud & API Tools

pip3 install flask fastapi uvicorn pip3 install grpcio grpcio-tools sudo apt install -y docker.io docker-compose

7. NVIDIA Extras (optional, powerful)

# Jetson.GPIO for hardware pins pip3 install Jetson.GPIO # DeepStream SDK (download from NVIDIA) # https://developer.nvidia.com/deepstream-getting-started

After Install

Reboot: sudo reboot
Test GPU: python3 -c "import torch; print(torch.cuda.is_available())"
Start JupyterLab: jupyter lab --ip=0.0.0.0 --port=8888

Jetson Nano install script you can save as setup_jetson.sh and run with bash setup_jetson.sh.

#!/bin/bash set -e echo "=== Updating System ===" sudo apt update && sudo apt upgrade -y sudo apt install -y build-essential cmake git curl wget unzip \ python3-pip python3-venv python3-dev \ libjpeg-dev zlib1g-dev echo "=== Upgrading pip ===" pip3 install --upgrade pip setuptools wheel virtualenv echo "=== Installing JupyterLab ===" pip3 install jupyterlab notebook echo "=== Installing Data Science Stack ===" pip3 install numpy pandas matplotlib seaborn scikit-learn pillow opencv-python echo "=== Installing AI Libraries ===" # PyTorch (you must replace the wheel URL with the latest for Jetson Nano / JetPack version) TORCH_WHL_URL="https://nvidia.box.com/shared/static/xxxxxx.whl" wget $TORCH_WHL_URL -O torch.whl pip3 install torch.whl pip3 install torchvision pip3 install tflite-runtime pip3 install onnxruntime echo "=== Installing IoT / Robotics ===" sudo apt install -y mosquitto mosquitto-clients pip3 install paho-mqtt pyserial # Uncomment if you want ROS (replace <distro> with melodic/noetic depending on Ubuntu) # sudo apt install -y ros-<distro>-ros-base echo "=== Installing Cloud / API Tools ===" pip3 install flask fastapi uvicorn grpcio grpcio-tools sudo apt install -y docker.io docker-compose echo "=== Installing NVIDIA Extras ===" pip3 install Jetson.GPIO echo "=== Setup Complete! Reboot recommended. ==="

Usage

Save the file:

nano setup_jetson.sh 
(Paste the script, save with CTRL+O, exit with CTRL+X)
Make it executable:

chmod +x setup_jetson.sh
Run it:

./setup_jetson.sh
Reboot when finished:

sudo reboot

Software to load on a NVIDIA Jetson Nano AI, robotics, and general development.