This package provides a list of rviz2
plugins for human-related
data visualisation. It is part of the ROS4HRI
ecosystem.
A plugin for visualising 2D information overlayed on a camera stream (ideally, the stream used to detect it). Currently, the plugin can visualise:
- Face bounding boxes
- Face landmarks
- Body bounding boxes
- 2D skeleton keypoints
- If not already available, add a
Displays panel
; - press the
Add
button at the bottom to istantiate a new plugin; - select
By topic
; - among the available topics, select the
Humans
plugin for the camera stream you are interested in; - once created the plugin, select which type of information you want to visualise (face bounging boxes, skeleton landmarks, etc.);
- enjoy!
A plugin for visualising the estimated 3D poses of the detected humans.
- If not already available, add a
Displays panel
; - press the
Add
button at the bottom to istantiate a new plugin; - select
By display type
; - select
Skeletons3D
; - enjoy!
A plugin for visualising the human-related TF frames. These are higly dynamical,
appearing and disappearing in a matter of seconds. Using the classic TF
plugin
would result in a crowded and chaotic frames visualisation. This plugin:
- looks over the detected faces and bodies;
- only displays the face and bodies TF frames for the currently detected bodies and faces;
- readily remove the TF frames for those bodies and faces that are no more tracked, avoiding the disappearing phase observed in the original TF frame for the non-updated frames. It is possible to select which human frames to visualise among:
- face frames;
- gaze frames;
- body frames;
- If not already available, add a
Displays panel
; - press the
Add
button at the bottom to istantiate a new plugin; - select
By display type
; - select
TF_HRI
; - enjoy!
To test the hri_rviz
plugins:
- Download
hri_face_detect
andhri_fullbody
; - build them;
- start an RGB camera stream;
- start face and body detection:
ros2 launch hri_face_detect face_detect.launch.py filtering_frame:=<camera_frame> rgb_camera:=<rgb_camera_stream_ns>
ros2 launch hri_fullbody hri_fullbody.launch.py rgb_camera:=<rgb_camera_stream_ns>
where<rgb_camera_stream_ns>
is the RGB camera stream namespace (e.g.,/camera/color
). This expects the raw RGB images to be published on<rgb_camera_stream_ns>/image_raw
. Check thelaunch
files parameters for different options.
- start the plugins as previously described.