This code contains algorithm to pop up a 3D plane model from images. Given a RGB and ground segmentation image, the algorithm detects edges and selecs the ground-wall boundary edges then pops up 3D point cloud.
Authors: Shichao Yang, Sebastian Scherer
Related Paper:
- Real-time 3D Scene Layout from a Single Image Using Convolutional Neural Networks, ICRA 2017, S. Yang, D. Maturana, S. Scherer PDF
- Pop-up SLAM: Semantic Monocular Plane SLAM for Low-texture Environments, IROS 2017, S. Yang, Y. Song, M. Kaess, S. Scherer PDF
If you use the code in your research work, please cite the above paper. Please do not hesitate to contact the authors if you have any further questions.
This code contains several ros packages. We test it in ROS indigo + Ubuntu 14.04. Create or use existing a ros workspace.
mkdir -p ~/popup_ws/src
cd ~/popup_ws/src
catkin_init_workspace
git clone [email protected]:shichaoy/pop_up_image.gitsh install_dependenices.shcd ~/popup_ws
catkin_makesource devel/setup.bash
roslaunch pop_up_wall pop_main_sample.launchYou will see point cloud in Rviz. Change the image id in the launch file to test more examples stored under pop_up_wall/data.