Skip to content
/ minisam Public
forked from dongjing3309/minisam

A general and flexible factor graph non-linear least square optimization framework

License

Notifications You must be signed in to change notification settings

cubao/minisam

 
 

Repository files navigation

make test_in_dev_container
conda install -y eigen sophus suitesparse
make build

miniSAM

Website: https://minisam.readthedocs.io/


miniSAM is an open-source C++/Python framework for solving factor graph based least squares problems. The APIs and implementation of miniSAM are heavily inspired and influenced by GTSAM, a famous factor graph framework, but miniSAM is a much more lightweight framework with

  • Full Python/NumPy API, which enables more agile development and easy binding with existing Python projects, and
  • A wide list of sparse linear solvers, including CUDA enabled sparse linear solvers.

miniSAM is developed by Jing Dong and Zhaoyang Lv. This work was initially started as final project of Math 6644 back to 2017, and mostly finished part-time when both authors were PhD students at College of Computing, Georgia Institute of Technology.

Mandatory Prerequisites

  • CMake 3.4+ (Ubuntu: sudo apt-get install cmake), compilation configuration tool.
  • Eigen 3.3.0+ (Ubuntu: sudo apt-get install libeigen3-dev), a C++ template library for linear algebra.

Optional Dependencies

  • Sophus, a C++ implementation of Lie Groups using Eigen. miniSAM uses Sophus for all SLAM/multi-view geometry functionalities.
  • Python 2.7/3.4+ to use miniSAM Python package.
  • SuiteSparse (Ubuntu: sudo apt-get install libsuitesparse-dev), a suite of sparse matrix algorithms. miniSAM has option to use CHOLMOD and SPQR sparse linear solvers.
  • CUDA 9.0+. miniSAM has option to use cuSOLVER Cholesky sparse linear solver.

Get Started

Please refer to https://minisam.readthedocs.io/install.html for more details.

To get and compile the library (on Ubuntu Linux):

$ git clone --recurse-submodules https://github.com/dongjing3309/minisam.git
$ mkdir build
$ cd build
$ cmake ..
$ make
$ make check  # optional, run unit tests

Tested Compatibility

The miniSAM library is designed to be cross-platform, should be compatible with any modern compiler which supports C++11. It has been tested on Ubuntu Linux and Windows for now.

  • Ubuntu: GCC 5.4+, Clang 3.8+
  • Windows: Visual C++ 2015.3+

Questions & Bug Reporting

Please use Github issue tracker for general questions and reporting bugs, before submitting an issue please have a look of this page.

Citing

If you use miniSAM in an academic context, please cite following publications:

@article{Dong19ppniv,
  author    = {Jing Dong and Zhaoyang Lv},
  title     = {mini{SAM}: A Flexible Factor Graph Non-linear Least Squares Optimization Framework},
  journal   = {CoRR},
  volume    = {abs/1909.00903},
  year      = {2019},
  url       = {http://arxiv.org/abs/1909.00903}
}

License

miniSAM is released under the BSD license, reproduced in the file LICENSE in this directory. Note that the linked sparse linear solvers have different licenses, see this page for details

About

A general and flexible factor graph non-linear least square optimization framework

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 94.0%
  • CMake 4.3%
  • Python 1.3%
  • Makefile 0.4%