AttentionClipper

What Is This

../../_images/attention_clipper.png

It retrieves sensor_msgs/CameraInfo and publish sensor_msgs/CameraInfo with ROI filled and retrieves sensor_msgs/PointCloud2 and publish pcl_msgs/PointIndices.

You can specify the pose and size of the interest bounding box and jsk_pcl/AttentionClipper returns ROI to see the object.

Subscribing Topic

  • ~input (sensor_msgs/CameraInfo)

    Original camera info. (You don’t need camera info when you only use pointcloud.)

  • ~input/points (sensor_msgs/PointCloud2)

    Original pointcloud.

  • ~input/pose (geometry_msgs/PoseStamped)

  • ~input/box (jsk_recognition_msgs/BoundingBox) Specify the pose of the bounding box. Timestamp will be ignored and camera info’s timestamp will be used. If you use ~input/box, you can change the size of attention region. There callbacks are only enabled if ~use_multiple_attention is false.

  • ~input/pose_array (geometry_msgs/PoseArray)

  • ~input/box_array (jsk_recognition_msgs/BoundingBoxArray) It’s an array version of ~input/pose and ~input/box. There callbacks are only enabled if ~use_multiple_attention is true.

Publishing Topic

  • ~output (sensor_msgs/CameraInfo)

    This camera info is same with ~input except for roi field. (only when ~input is published.)

  • ~output/box_array (jsk_recognition_msgs/BoundingBoxArray)

    Array of bounding boxes representing the interest regions.

  • ~output/mask (sensor_msgs/Image)

    Mask image to mask the regions of specified interest. (only when ~input is published.)

  • ~output/point_indices (pcl_msgs/PointIndices)

    Indices of ~input/points which are inside of interest regions.

  • ~output/cluster_point_indices (jsk_recognition_msgs/ClusterPointIndices)

    Cluster point indices of ~input/points which are inside of interest regions.

Parameter

  • ~use_multiple_attention (Boolean, default: False)

    If you want to enable multiple attentions, please set this variable True.

  • ~dimension_x (Double, default: 0.1)

  • ~dimension_y (Double, default: 0.1)

  • ~dimension_z (Double, default: 0.1)

    Size of bounding box. Available only if ~use_multiple_attention is false.

  • ~frame_id (String, default: base_link)

    Frame id of attention region. Available only if ~use_multiple_attention is false.

  • ~initial_pos (Array of double, default: None):

    Initial pose of interesting region. Available only if ~use_multiple_attention is false.

  • ~initial_rot (Array of double, default: None):

    Initial orientation of interesting region. The value should be represented in [roll, pitch, yaw]. Available only if ~use_multiple_attention is false.

  • ~initial_pos_list (Array of array of double, default: None)

  • ~initial_rot_list (Array of array of double, default: None)

  • ~frame_id_list (Array of string, default: None)

  • ~dimensions (Array of array of double, default: None)

  • ~prefixes (Array of string, default: None)

    Position, Rotation, frame id ,prefix and Dimensions of multiple attention regions respectively. ~iniital_pos_list should follow [[x, y, z], ...], ~initial_rot_list should follow [[rx, ry, rz], ...] and ~dimensions should follow [[x, y, z], ...]. ~prefixes [prefix1, prefix2, ...]. These prefixes will add to the /point_indices and advertise Available only if ~use_multiple_attention is true.

  • ~negative (Boolean, default: False)

    Publish points which are not inside of attention regions if this parameter is true.

Sample

roslaunch jsk_pcl_ros sample_attention_clipper.launch