OpenVINO™ Object Detection Tutorial

This tutorial is an example for understanding the utilization of the ROS 2 OpenVINO™ node. It outlines the steps for installing the node and executing the object detection model. Object detection is performed using the OpenVINO™ toolkit. The node is configured to accept dynamically device parameters (NPU, GPU, or CPU) to specify which inference engine should be used.

Install OpenVINO™ tutorial packages

sudo apt install ros-humble-object-detection-tutorial
Copy to clipboard

Run Demo with Image Input

Run one of the following commands to launch the object detection node with a specific inference engine:

  • GPU inference engine

    ros2 launch object_detection_tutorial openvino_object_detection.launch.py device:=GPU
    Copy to clipboard

  • CPU inference engine

    ros2 launch object_detection_tutorial openvino_object_detection.launch.py device:=CPU
    Copy to clipboard

  • NPU inference engine

    ros2 launch object_detection_tutorial openvino_object_detection.launch.py device:=NPU
    Copy to clipboard

Note

If no device is specified, the GPU is selected by default as an inference engine.

Once the tutorial is started, the mobilenetssd model is downloaded, converted into IR files, and the inference process begins.

Expected Output

../../../../_images/Object_detection.png

To close this application, type Ctrl-c in the terminal where you ran the launch script.

Troubleshooting

For general robot issues, go to: Troubleshooting for Robotics SDK Tutorials .