benchun123 / cuboid_slam_with_bbox_constraints

open source code for cuboid slam with bbox constraints

3d Object Estimation with Bounding Box Constraints

Related Paper:

• Benchun Zhou, Aibo Wang, Jan-Felix Klein, Furmans Kai. Object Detection and Mapping with Bounding Box Constraints[C]//2021 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI). IEEE, 2021. [Link] [PDF] [Slide] [Youtube] [Bilibili]

• If you use the code in your academic work, please cite the above paper.

this code is just for cabinet dataset

directory

in folder /cmake_det_slam, we can find the cmake code to detect object and use object for SLAM

in folder /ros_det_slam, we add visulazation for SLAM

in folder /data, we save some neccessary images or files.

how to run the project

1. data folder should contain:
   -raw_imgs folder with images and rgb.txt
   -depth_imgs folder with images and depth.txt
   -bbox_2d_yolov2 folder with 2d bbox result and yolov3_bbox.txt
   -TUM3.yaml # calibration
   -truth_cam_poses.txt # ground truth for camera -truth_objects.txt # ground truth for object (measured from point cloud)

2. to build essential thirdparty library

  cd cmake_det_slam/Thirdpary  
  cd line_lbd  
  mkdir build  
  cd build  
  cmake ..  
  make  

it will output a library "libline_lbd_lib.so" in line_lbd/lib

  cd cmake_det_slam/Thirdpary  
  cd ticioc_profiler  
  mkdir build  
  cd build  
  cmake ..  
  make  

it will output a library "libtictoc_profiler.so" in ticioc_profiler/lib

3. to build and run cmake code:

  cd cmake_det_slam  
  mkdir build  
  cd build  
  cmake ..  
  make  
  ./bbox_cabinet_det_node ../../data #path_to_cabinet data

it can output "online_camera.txt" and "online_cube.txt" to save the result.

4. to run ros code:
   firstly, you also need to build essential thirdparty library in (2) secondly, you need to build g2o in thirdparty library

  cd ros_det_slam/Thirdpary  
  cd g2o  
  mkdir build  
  cd build  
  cmake ..  
  make

it will output a library "libg2o.so" in g2o/lib

thridly, build and run visulazation code:

  cd ros_det_slam/src  
  catkin_init_workspace  
  cd ..  
  catkin_make
  source devel/setup.bash    
  roslaunch object_slam_bbox object_slam_example.launch   

Acknowledgement

• The code referred to several open-source SLAM codes, thanks to their great work: ORBSLAM, CubeSLAM.