README.md

Shape map 3-D: Efficient shape mapping through dense touch and vision

    

   

Base library for visuo-tactile shape mapping with gelsight and depth camera. For more information, consider our work Efficient shape mapping through dense touch and vision.

Roadmap

The library is still being actively updated, please check back soon for a stable version!

• Initial code upload
• Data upload
• Compilation fixes
• Enhanced documentation

Folder structure

• matlab/: Executable and utility scripts for mapping
  • shape-map/: 3-D and 2-D mapping scripts for both simulated and real-world data
  • utils/: helpers scripts
• gpfactor/: cpp header wrapped with the gtsam + MATLAB library

Requirements

• gtsam
• MATLAB

Other dependencies

• spatialmath-matlab

Third-party scripts

• splitFV
• stlwrite

Compilation

• GTSAM: Checkout gtsam 4.0.2 master branch (not develop):
  • Using the default CMakeLists without modification

    cmake -DGTSAM_USE_SYSTEM_EIGEN=ON -DGTSAM_WITH_TBB=OFF -DGTSAM_INSTALL_MATLAB_TOOLBOX=ON ..
    sudo make install -j
    

  • MATLAB toolbox check
    • Open MATLAB
    • add addpath /usr/local/gtsam_toolbox to the startup.m
    • Test gtsamExamples
• C++ header compilation:

  mkdir build 
  cd build
  cmake -DGPFACTOR_BUILD_TOOLBOX:OPTION=ON -DCMAKE_BUILD_TYPE=Release -DGTSAM_TOOLBOX_INSTALL_PATH:PATH=/usr/local/gtsam_toolbox ..
  sudo make install 
  

  • sudo ldconfig in terminal
  • open MATLAB here
  • Run .m files

Citation

Feel free to use the library as you please. If you find it helpful, please consider referencing:

@misc{suresh2021efficient,
      title={Efficient shape mapping through dense touch and vision}, 
      author={Sudharshan Suresh and Zilin Si and Joshua G. Mangelson and Wenzhen Yuan and Michael Kaess},
      year={2021},
      eprint={2109.09884},
      archivePrefix={arXiv},
      primaryClass={cs.RO}
      }