The “zed_description” Package
The “zed_description” Package
The zed_description package provides the 3D models (URDF/xacro and meshes) of the ZED camera family, used by the ZED ROS 2 Wrapper and the ZED ROS 2 Examples to publish the static TF tree of the camera and to display it in RViz 2. The package is maintained in the zed-ros2-description repository.
The zed_description package does not depend on CUDA or the ZED SDK. You can install it on any ROS 2 machine — even without an NVIDIA GPU — to visualize a ZED model or build a robot URDF that includes a ZED camera.
Installation
Binary installation
where <ros2-distro> is your ROS 2 distribution (e.g. humble or jazzy).
Build from source
Supported camera models
The package ships a mesh and the URDF parameters (baseline, optical offsets) for each supported model. The camera_model argument accepts one of the following values:
zed, zedm, zed2, zed2i, zedx, zedxm, zedxnano, zedxhdr, zedxhdrmini, zedxhdrmax, zedxonegs, zedxones, zedxone4k, zedxonehdr, and virtual (for custom stereo rigs).
Visualizing a camera model
You can preview the 3D model of any supported camera in RViz 2 without starting a ZED node:
This launch file is meant for URDF debugging and visualization only. It does not start a ZED node and therefore does not stream any camera data. To run an actual camera, use the ZED Stereo Node or ZED Monocular Node launch files.
Using the description in your own URDF
The model is generated from a xacro macro (zed_macro.urdf.xacro) that you can include in your own robot description to attach a ZED camera to your platform:
The macro accepts the following arguments:
- name: name of the camera, used as a prefix for all the generated frames (default
zed). - model: camera model, one of the supported values listed above.
- custom_baseline: distance in meters between the left and right optical centers. Only used with
model:=virtualto describe a custom stereo rig. - enable_gnss: if
true, adds agnss_linkframe for the GNSS antenna. Used by the Geo Tracking module. The antenna position relative to the camera mount point is set through thegnss_x,gnss_y, andgnss_zarguments.
The complete TF tree generated by this description, including the optical frames of each sensor, is detailed in the Getting Started and Robot Integration pages.

