ORB-SLAM3 for Windows

ORB-SLAM3 Project for Windows Platform (espacially Visual Studio)
We have tested the library in Windows 10, Visual Studio 2015.

Reference

• ORB-SLAM3 project: https://github.com/UZ-SLAMLab/ORB_SLAM3
• ORB-SLAM2 for windows: https://github.com/Phylliida/orbslam-windows
• ORB-SLAM2 for windows_2: https://github.com/phdsky/ORBSLAM24Windows

To do...

Update from ORB-SLAM3 V0.2 to V0.3

Prerequisites

1. Visual Studio: Tested with VS2015

2. Cmake: Required at least 2.8. (Tested with 3.14.7)

3. OpenCV: Required at least 3.0. (Tested with 3.4.1)

4. glew (for pangolin): Tested with 2.1.0

5. boost: Tested with 1.67.0: https://www.boost.org/users/history/version_1_67_0.html

Build

eigen

DBoW2

• Open cmake-gui and make a 'build' directory in ORB-SLAM3forWindows/Thirdparty/DBoW2
• Click Configure and select Visual Studio 14 2015 x64.
• Click Generate and open the resulting project in the build directory.
• Change build type to Release.
• Right click on DBoW2 project -> Properties -> General -> Target Extension: .lib, and Configure Type: Static Library (.lib)
• DBoW2 project -> Properties -> C/C++ Tab -> Code Generation -> Runtime Library: Multi-threaded(/MT)
• Build ALL_BUILD. Finish(DBoW2.lib).

g2o

• Open cmake-gui and Make a 'build' directory in ORB-SLAM3forWindows/Thirdparty/g2o
• Click Configure and select Visual Studio 14 2015 x64.
• Click Generate and open the resulting project in the build directory.
• Change build type to Release.
• Right click on g2o project -> Properties -> General -> Target Extension: .lib, and Configure Type: Static Library (.lib)
• g2o project -> Properties -> C/C++ Tab -> Code Generation -> Runtime Library: Multi-threaded(/MT)
• g2o project -> Properties -> C/C++ Tab -> Preprocessor -> Preprocessor Definitions: add a 'WINDOWS'.
• Build ALL_BUILD. Finish(g2o.lib).

Pangolin

• Open cmake-gui and Make a 'build' directory in ORB-SLAM3forWindows/Thirdparty/Pangolin
• Click Configure and select Visual Studio 14 2015 x64.
• cmake error: could not find git for clone of ... : Disable BUILD_EXTERN_GLEW, LIBJPEG, LIBPNG and input value directly (used GLEW only)
• If window says Configuring Done, it's OK
• Click Generate and open the resulting project in the build directory.
• Change build type to Release.
• pangolin project -> Properties -> C/C++ Tab -> Code Generation -> Runtime Library: Multi-threaded(/MT)
• Build ALL_BUILD. (ignore error "cannot open input file 'pthread.lib') Finish(pangolin.lib).

ORB-SLAM3 for Windows

• Open cmake-gui and make a 'build' directory in ORB-SLAM3forWindows
• Click Configure and select Visual Studio 14 2015 x64.
• Click Generate and open the resulting project in the build directory.
• Change build type to Release.
• Setting Boost Library (include, dir, ...)
• Right click on ORB-SLAM3 project -> Properties -> General -> Target Extension: .lib, and Configure Type: Static Library (.lib)
• ORB-SLAM3 project -> Properties -> C/C++ Tab -> Code Generation -> Runtime Library: Multi-threaded(/MT)
• ORB-SLAM3 project -> Properties -> C/C++ Tab -> Preprocessor -> Preprocessor Definitions: check 'COMPILEDWITHC11'.
• Build ORB-SLAM3 (not ALL_BUILD). Finish(ORB_SLAM3.lib)

Test Program

We have tested the following projects: mono_euroc, mono_inertial_euroc, stereo_euroc, stereo_inertial_euroc

example: The EuRoC MAV Dataset: monocular-inertial

• Open mono_inertial_euroc directory in ORB-SLAM3forWindows.
• Change build type to Release.
• Setting Boost Library (include, dir, ...)
• Right click on mono_inertial_euroc project -> Properties -> C/C++ Tab -> Code Generation -> Runtime Library: Multi-threaded(/MT)
• mono_inertial_euroc project -> Properties -> Linker -> Advanced -> Delete import library.
• Define 'usleep' function in mono_inertial_euroc.cc
• Build mono_inertial_euroc. Finish(mono_inertial_euroc.exe).
• Visual Studio 0xC00000FD: Stack overflow error: Right click on the project -> Properties -> Linker -> System -> Stack Reserve Size: 8388608(8MB) or more (the default stack size is 1MB = 1048576)
• You will find the test program in ORB-SLAM3forWindows\Examples\Monocular-Inertial\Release.
• Download the EuRoC MAV Dataset: https://projects.asl.ethz.ch/datasets/doku.php?id=kmavvisualinertialdatasets
• Convert ORBvoc.txt.tar.gz -> ORBvoc.txt
• Execute the following script:
  mono_inertial_euroc.exe (path_to_vocabulary) (path_to_settings) (path_to_image_folder_1) (path_to_times_file_1) (path_to_image_folder_2) (path_to_times_file_2) ... (path_to_image_folder_N) (path_to_times_file_N)
• The result of Machine Hall 01 Dataset (monocular-inertial):