How to launch the different components for person detection¶

This how-to describes how to test the hri_person_manager node in your local environment to familiarise yourself with the ROS4HRI pipeline. The node aggregates information about faces, bodies and voices in a single structure, person, which facilitates the management of the human features detected around the robot.

Prerequisites¶

You will need the hri_face_detect and hri_face_identification ROS nodes. These nodes will feed hri_person_manager. They are pre-installed in the PAL Docker images.

Check Developing with the PAL Developer Docker image if you have not yet installed your Docker image.
You will also need a webcam and a calibration file for the camera. If you don’t have one, here you can find a standard calibration file sample that you can temporarily use for this tutorial. If you are using this calibration file, do not expect the face pose estimation to be precise.

Testing¶

In terminal 1, start gscam:

ros2 run gscam gscam_node --ros-args -p gscam_config:='v4l2src device=/dev/video0 ! video/x-raw,framerate=30/1 ! videoconvert' \
                                     -p use_sensor_data_qos:=True \
                                     -p camera_name:=camera \
                                     -p frame_id:=camera \
                                     -p camera_info_url:="file:///<calibration_file_location>"

Note

You must provide a calibration file for your camera (like this generic file for webcams), otherwise hri_face_detect will not be able to publish the TF frames of the faces (note that the name of the camera in the calibration file should be default_cam, as I will assume this name for the camera TF frame).

Without calibration, hri_person_manager will still partially work, but for example, proximal zones will not be published, and the person TF frames will not be available.

Note

Depending on your setup, your camera might be avaiable under a different device name (e.g., /dev/video1). If so, change the device parameter accordingly in the command above.

In terminal 2, start hri_face_detect:

First, configure the remappings for the node: by default, gscam publishes the camera image on /camera/image_raw and the camera info on /camera/camera_info. We need to tell hri_face_detect about it:

mkdir -p $HOME/.pal/config
nano $HOME/.pal/config/ros4hri-tutorials.yml

Then, paste the following content:

/hri_face_detect:
   remappings:
      image: /camera/image_raw
      camera_info: /camera/camera_info

Press Ctrl+O to save, then Ctrl+X to exit.

Then, launch the node:

$ ros2 launch hri_face_detect face_detect.launch.py

In terminal 3, start hri_face_identification:

$ ros2 launch hri_face_identification face_identification.launch.py

In terminal 4, publish a static (identity) transform between the camera and /map:

$ ros2 run tf2_ros static_transform_publisher --frame-id map --child-frame-id default_cam

In terminal 5, start the hri_person_manager node:

$ ros2 launch hri_person_manager person_manager.launch.py

In terminal 6, check the output of /humans/persons/tracked, /humans/persons/in_{personal,social,public}_space, /humans/persons/<id>/*…

For instance, you should see your own person ID:

$ ros2 topic echo /humans/persons/in_personal_space

Optionally, you can display the face - person association graph:
- In terminal 7, run:
```
$ ros2 run hri_person_manager show_humans_graph
```
- In terminal 8, view the output:
```
$ evince /tmp/graph.pdf
```

That’s all! You can now start developing your own code using the available libraries in C++ and python, or continue playing with other tutorials in the 👥 Social perception main page.

How to launch the different components for person detection¶

Prerequisites¶

Testing¶

See also¶