6D Object Pose Estimation (D6.4 ILIAD)

This contains a brief guide how to install / run the ROS-based simplified Object-RPE developed at ORU in D6.4 ILIAD project.

The set of 11 objects in our warehouse dataset.

Installation and compile the source

The tools require full ROS installation. The installation assumes you have Ubuntu 16.04 LTS [ROS Kinetic]

1. Clone the repository and switch to iliad branch

   $ https://github.com/hoangcuongbk80/Object-RPE.git
   $ git checkout iliad
   

2. ROS

   $ cd ~/catkin_ws
   $ catkin_make install
   

3. Segmentation here
4. 3D mapping here
5. 6D object pose estimation here

Testing

1. Download data folder here and copy to ~/catkin_ws/src/Object-RPE

2. To test segmenation module:

   $ cd ~/catkin_ws/src/Object_RPE/Mask-RCNN/samples/warehouse
   $ python3 eval.py
   

   It will read rgb and ground-truth images from /rgb and gt/ folders in .../Object-RPE/data then save results (mask and accuracy.txt file) into .../Object-RPE/data/mask

   

3. To test 3D mapping module:

   $ cd ~/catkin_ws/src/Object-RPE/obj_pose_est/mapping/app
   $ ./obj_pose_est/mapping/app/build/mapping -l data/ 100
   

   

4. To test pose estimation module (Video):

   $ roscore
   

   In an other terminal:

   $ rosrun obj_pose_est ObjectRPE_srv.py
   

   In an other terminal:

   $ roslaunch obj_pose_est launch_object_rpe.launch
   

   

How to operate the system?

$ roscore
$ rosrun obj_pose_est ObjectRPE_srv.py
$ roslaunch obj_pose_est launch_rpe_cam.launch
$ roslaunch openni2_launch openni2.launch