ROS2 CVNode

Copyright (c) 2022-2024 Antmicro

CVNode is a ROS2 node designed to facilitate integration of computer vision algorithms into inference testing infrastructure.

Overview

CVNode introduces the CVNodeBase class, which serves as a foundational building block for creating computer vision nodes within ROS2 projects. This base class provides essential functionality for running ROS2 nodes. It only requires implementation of the following abstract methods to set up your computer vision algorithm:

• prepare - responsible for preparing the computer vision algorithm for inference (e.g. load the model).
• run_inference - responsible for running the computer vision algorithm on a vector of input images.
• cleanup - responsible for cleaning up the computer vision algorithm after inference.

CVNode offers both C++ and Python implementations of the CVNodeBase class, enabling choice of computer vision algorithm development language.

Building CVNode

Project dependencies:

• ROS2 Humble
• kenning_computer_vision_msgs
• cv_bridge
• OpenCV

The CVNodeBase class is located in basecvnode target, which is a shared library. To build the basecvnode target, run the following command from the root of your ROS2 workspace:

colcon build --packages-select cvnode_base

This will build libbasecvnode.so, which can later be used as a dependency for your computer vision node.

For a usage sample of the CVNodeBase class, see the MaskRCNN demo in the examples/mask_rcnn directory. You can also explore the source code of implemented nodes in the cvnode_base/nodes (Python implementations) and include/cvnode_base/nodes (C++ implementations) directories.