people_mask_publisher.py

What is this?

../../_images/people_mask_publisher.png

Create people mask image according to people pose information.

Subscribing Topic

  • ~input (sensor_msgs/Image)

    Input image.

  • ~input/pose (jsk_recognition_msgs/PeoplePoseArray)

    Input people pose array.

Publishing Topic

  • ~output (sensor_msgs/Image)

    People mask image.

  • ~debug/output (sensor_msgs/Image)

    Debug image which has visualized rectangle of the masked limb part.

Parameters

  • ~person_indices (Int, Default: -1)

    Which person to create mask of. -1 represents creating masks of all people.

  • ~limb_part (String or List of String, Default: all)

    Which limb part to create mask of.

    RHand, LHand, Nose and all is allowed.

  • ~arms_score_threshold (Float, Default: 0.25)

    Threshold of arm score.

    Used when limb_part includes RHand or LHand.

  • ~hand_ratio (Float, Default: 0.33)

    Ratio that is used to predict hand position from elbow and wrist positions.

    Used when limb_part includes RHand or LHand.

  • ~hand_width_ratio (Float, Default: 0.8)

    Ratio that is used to predict hand region from arm length.

    Used when limb_part includes RHand or LHand.

  • ~face_ratio (Float, Default: 0.6)

    Ratio that is used to predict face position from nose and neck positions.

    Used when limb_part includes Nose.

  • ~face_shoulder_ratio (Float, Default: 0.5)

    Ratio that is used to predict face position from shoulder positions.

    Used when limb_part includes Nose.

  • face_width_margin_ratio (Float, Default: 1.3)

    Ratio that is used to decide face width margin.

    Used when limb_part includes Nose.

  • ~approximate_sync (Bool, Default: False)

    Use approximate synchronization policy.

  • ~queue_size (Int, Default: 10)

    Queue size for synchronization.

  • ~slop (Float, Default: 0.1)

    Slop for approximate sync.

Example

roslaunch jsk_perception sample_people_pose_estimation_2d.launch GPU:=0 LIMB_PART:='RHand'
roslaunch jsk_perception sample_people_pose_estimation_2d.launch GPU:=0 LIMB_PART:='Nose'