This package contains the code for the Viewpoint Pose Tree introduced by Sankaran et al in ICRA-13 and TRO-14.

Database Download
-----------------
1) First Download the object models and test scenes in the database directory located at 
http://clmc.usc.edu/~bharath/viewpoint_pose_tree/ to a database directory inside the viewpoint_pose_tree repository you have cloned from git


Installation and Compiling
--------------------------
1) First create a catkin workspace and clone the repository into this workspace
   
   source /opt/ros/groovy/setup.bash
   cd ~/
   mkdir VIEWPOINT_POSETREE && cd VIEWPOINT_POSETREE
   mkdir src && cd src
   catkin_init_workspace

2) Build the workspace

   cd ..
   catkin_make
   source devel/setup.bash
   # you can add the source line to your bash with an easy alias like this
   # add the following line to your bashrc
   alias groovy_viewpoint='source ~/VIEWPOINT_POSETREE/devel/setup.bash'

3) Clone and compile the code

   cd src
   git clone https://github.com/bsankaran/viewpoint_pose_tree.git
   cd ..
   catkin_make

Database Download
-----------------
1) First Download the object models and test scenes in the database directory located at
http://clmc.usc.edu/~bharath/viewpoint_pose_tree/ to a database directory inside the viewpoint_pose_tree repository you have cloned from git


Running the Code and Testing Functionality
------------------------------------------
1) Building the database of objects with partial view

# PRECONDITIONS: 1.1) Ensure the database directory is in the right folder path, i.e inside the viewpoint_pose_tree root directory
#                1.2) Create a directory called cloud_data in the database directory 
#                     mkdir cloud_data
#                1.3) Make sure you install the GNU parallel package using the instructions provided in vision_module/script/extract_tree_train_pcds
#                1.4) Ensure that extract_tree_train_pcds is an executable, you can do this by running
#                     chmod 777 extract_tree_train_pcds
#                1.5) Run the bash script extract_tree_train_pcds in the vision_module/scripts folder 
#                     ./extract_tree_train_pcds
#                1.6) Ensure that the views have been extarcted properly, they should be located in the viewpoint_pose_tree/database/cloud_data folder
#                     The extracted views can be visualized using pcd_viewer

Once you've statisfied the preconditions you can run the following command to build your VP-Tree
      
                 roslaunch vision_module build_tree_database.launch

# Comment: Now you can get a cup of coffee/dinner or call it a night at this point, since it takes a while to build the tree. The time taken to build the tree depends on the object models you have used,
# The number of cores your computer has and the density of the models. On a slow computer this can take an entire evening.
# You can choose to use the models we provide or put your own models into database/models you'll have to edit the extract_tree_train_pcds file appropriately

Once you the tree is built, the process automatically terminates and the tree related files, clouds.documents, clouds.tree, clouds.weights can be found in vision_module/data/ folder

2) To test and run the virtual kinect launch the launch file located in the virtual kinect pkg
     
      roslaunch virtual_kinect_pkg test_virtual_kinect.launch

# NOTE: You can uncomment the rviz visualization to see the output

3) To use this code to write your own VP-TREE analyzer, check the source code in the test_vp_tree pkg. It contains a simple_point_cloud_analyzer.cpp that uses the tabletop analyzer and vp_tree for preprocessing and analyzing the data using the VP-Tree. This code can be run using the launch file located in the same package

     roslaunch test_vp_tree test_vp_tree.launch

# NOTE: Remember to adapt the parameters in the launch file with the correct topics and frames used by your robot.

4) To simply test the VP tree you can run the following command
  
   rosrun vision_module score_pcd <your_view>.pcd

   OR

  rosrun vision_module score_pcd -sp /path/to/save/score.txt <your_view>.pcd