openFABMAP

Open Source C++ Code for the FAB-MAP Algorithm

St. Lucia Multiple Times of Day and Other Datasets

See the Wiki for tips!

@inproceedings{
   author = {Glover, A. and Maddern, W. and Warren, M. and Reid, S. and Milford, M. and Wyeth, G.},
   title = {OpenFABMAP: An Open Source Toolbox for Appearance-based Loop Closure Detection},
   booktitle = {The International Conference on Robotics and Automation},
   address = {St Paul, Minnesota},
   publisher = {IEEE},
   year = {2011}
}

OpenFABMAP Glover et. al. 2012 is an open-source, OpenCV-only dependent, version of the popular Fast Appearance-based Mapping (FAB-MAP) algorithm [Cummins & Newman 2008](http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5509547&tag=1 Glover et al. 2010). OpenFABMAP was developed from the ground-up following FAB-MAP publications. The original FAB-MAP algorithm is now also open-source but requires alternative project dependencies.

FAB-MAP is a Simultaneous Localisation and Mapping algorithm which operates solely in appearance space. FAB-MAP performs location matching between places that have been visited within the world as well as providing a measure of the probability of being at a new, previously unvisited location. Camera images form the sole input to the system, from which OpenCV's feature extraction methods are used to develop bag-of-words representations for the Bayesian comparison technique.

The code has implementations of

• Feature Detection, Feature Extraction, and Bag-of-words models using OpenCV
• Chow-Liu tree implementation
• FAB-MAP v1.0 [Cummins & Newman 2008](http://ijr.sagepub.com/content/27/6/647.short Cummins & Newman 2008)
• FAB-MAP v1.0 using a Look-up-table for improved computation speed
• FAB-MAP with Fast-Bailout Cummins & Newman 2010
• FAB-MAP v2.0 Cummins & Newman 2010

An overview of OpenFABMAP Glover et. al. 2012 or the original implementation/use Glover et al. 2010.

As of the latest version, openFABMAP is dependent solely on OpenCV 2.3 or higher. The project has a CMake build environment for general use on both Linux and Windows systems. OpenFABMAP is also designed to integrate with ROS. See the CyPhy-ROS page for a package that has implemented openFABMAP as a ROS node.

The original googlecode project page was here

Installation

Linux (g++)

1. install cmake sudo apt install cmake
2. install opencv sudo apt install libopencv-dev
3. get the openFABMAP code git clone https://github.com/arrenglover/openfabmap.git
4. make the build directory mkdir openfabmap/build && cd openfabmap/build
5. use cmake to compile the makefile cmake ... note: the output will tell you which version of opencv you are using and if you are using the "non-free" modules
6. make the project make
7. view/modify the settings file gedit ../samples/settings.yml
8. run the command line tool bin/openFABMAPcli -s ../samples/settings.yml

OpenCV non-free for OpenCV 3.4

1. clone opencv_contrib
2. clone opencv repository
3. checkout version 3.4
4. mkdir build && cd build
5. cmake .. -DOPENCV_EXTRA_MODULES_PATH='path_to/opencv_contrib/modules' -DBUILD_opencv_xfeatures2d=ON -DOPENCV_ENABLE_NONFREE=ON
6. make

Windows (Visual Studio 2008)

1. install openCV2.3
2. install cmake
3. open the cmake gui, specify the source directory (the directory this README is in), a build directory for the code, and click configure
4. you may have to specify the location of opencv2.3 in UngroupedEntries->OPENCV_PATH.
5. click configure in the cmake gui again
6. click generate
7. open the visual studio solution, for default running right-click on openFABMAPexe project and select 'Set as StartUp project'. Compile openFABMAP within Visual studio.
8. add required .dll files from openCV2.3 to your build/bin directory (respective debug versions for debug mode).
9. you also may need an extra tbb .dll due to OpenCV bug which can be downloaded here
10. under openFABMAPcli->properties->Debugging->command arguments specify the path to the settings file (e.g. "-s samples\settings.yml")
11. Alter the settings file for your data
12. run exampleopenFABMAP in your build/bin directory (respective debug versions for debug mode).

Contributors

• Arren GLover
• Will Maddern
• Kirk MacTavish