stereolabs/zed-ros2-wrapper

What it solves

This project provides a ROS 2 wrapper that allows developers to integrate Stereolabs ZED cameras into the Robot Operating System 2 (ROS 2) ecosystem. It bridges the gap between the ZED SDK and ROS 2, enabling robots to access high-fidelity spatial data and AI-driven perception capabilities as standard ROS topics.

How it works

The wrapper acts as a node that interfaces with the ZED SDK and publishes various data streams to the ROS 2 graph. It supports multiple ZED camera models and can be configured via YAML files. It leverages "manual composition" to start nodes and integrates with NVIDIA Isaac ROS via NITROS for accelerated data streaming. For advanced perception, it can run object detection (including custom YOLO-like ONNX models) and body tracking directly through the ZED SDK and expose the results to ROS.

Who it’s for

Robotics developers and researchers using ZED cameras who need to integrate depth sensing, positional tracking, and AI perception into a ROS 2-based robot control system.

Highlights

• Comprehensive Data Streams: Provides access to rectified color/grayscale images, depth maps, 3D point clouds, and sensor data.
• Spatial Intelligence: Includes positional tracking, mapping, and optional GNSS data fusion for precise localization.
• AI Perception: Supports built-in object detection, human body tracking, and the ability to use custom YOLO-like ONNX models for specific detection tasks.
• Simulation Support: Compatible with NVIDIA Omniverse Isaac Sim for testing in virtual environments.
• Hardware Acceleration: Optimized for NVIDIA platforms with integration for Isaac ROS and NITROS.