Jetson Containers is an open-source, modular container build system that packages the latest AI and machine learning software for NVIDIA Jetson edge devices. Maintained by Dustin Franklin and the community, it has collected more than 4,800 stars on GitHub and is updated continuously, including in 2026.
Getting modern ML frameworks running on Jetson’s ARM platform with JetPack and L4T has historically been painful, because versions of CUDA, PyTorch, TensorFlow and system libraries must all line up. Jetson Containers solves this with a modular build system that assembles Docker images with matched versions of popular packages: PyTorch, TensorFlow, JAX, ONNX Runtime, TensorRT ecosystem tools, DeepStream, ROS and ROS 2, JupyterLab and many more. The project also runs a PyPI index with prebuilt Jetson wheels, so developers can often install packages directly without building containers at all. It is closely tied to the Jetson AI Lab resources for edge AI and robotics.
In practice, the build system inspects the JetPack and L4T versions of the target device and resolves compatible package versions automatically. That removes much of the trial and error normally involved in assembling a working deep-learning environment on Jetson hardware, where a single mismatched library version can break an entire deployment pipeline.
Jetson Containers is free and open source. The build system, package definitions and prebuilt wheels are available from GitHub and the Jetson AI Lab package index at no cost.
Jetson Containers is for embedded and edge AI developers deploying models on NVIDIA Jetson hardware, from hobbyist robots to commercial products. Anyone who has struggled with framework installation on JetPack will save hours by starting from these tested container definitions instead of compiling dependencies by hand.