ros package for object detection.

rosdep install -y -r --from-paths --ignore-src .
catkin build neatness_estimator

• input

1. ~input_image (sensor_msgs/Image)
2. ~input_cloud (sensor_msgs/PointCloud2)

• output

There are topics only often used

1. /labeled_bounding_box_publisher/labeled_instance_boxes (jsk_recognition_msgs/BoundingBoxArray)
2. /labeled_bounding_box_publisher_aligned/labeled_instance_boxes (jsk_recognition_msgs/BoundingBoxArray)
3. /labeled_bounding_box_publisher_aligned/labeled_cluster_boxes (jsk_recognition_msgs/BoundingBoxArray)

roslaunch neatness_estimator mask_rcnn_clustering2.launch

• input

1. ~input_cluster_indices (jsk_recognition_msgs/ClusterPointIndices)
2. ~input_point_cloud (sensor_msgs/PointCloud2)

• output

1. ~output_indices (jsk_recognition_msgs/ClusterPointIndices)

roslaunch neatness_estimator multi_euclidean_clustering.launch

Each bounding boxes results are published as /labeled_bounding_box_publisher/labeled_instance_boxes topic.
Boxes pose are aligned by pca coordinates.

1. Same class surrounding cluster bounding boxes aligned by target frame
   • /labeled_bounding_box_publisher_aligned/labeled_cluster_boxes
2. Each class instance bounding boxes aligned by target frame
   • /labeled_bounding_box_publisher_aligned/labeled_instance_boxes

under developing

under developing

roslaunch neatness_estimator distance_estimator_module.launch

• get distance between two scenes

rosservice call /estimation_module_interface/call ${features srv message}

• get distance between same class items feature work

feature work