Interpretation of RosTopic

[NorrieDS]

Thank you very much for building such an excellent toolkit! I am currently using its object segmentation in my project and have successfully launched openvino_node for segmentation with an HRNet model. Now, I would like to use the published segmentation results in my custom node for further processes and checked the contents of the topic:

ros2 topic echo /ros2_openvino_toolkit/segmented_objects

header: stamp: sec: 47 nanosec: 634000000 frame_id: camera_color_optical_frame objects_vector:

• object_name: '' probability: -1.0 roi: x_offset: 0 y_offset: 0 height: 320 width: 320 do_rectify: false mask_array:
  • -1.1778081748465806e-38
  • 4.015655517578125
  • -5.923568868393867e-39
  • -1.1778081748465806e-38
  • 4.015655517578125
  • -5.923568868393867e-39
  • -1.1778081748465806e-38
  • 4.015655517578125
  • -5.923568868393867e-39
  • -1.1778081748465806e-38
  • 4.015655517578125
  • -5.923568868393867e-39 ...

Having expected an array ascribing each pixel to a class, I do not understand what this data means. "probability: -1.0" in the header looks mysterious to me as well. Could anyone explain what the topic represents and/or show me how I can convert this to a segmentation map in the cv::Mat format? Thank you very much for your help!

[NorrieDS]

I noticed that I should subscribe to /ros2_openvino_toolkit/image_rviz rather.