Installation instructions

SISPO can be setup with Linux and Windows. The default case used in this description is a Windows setup. It is recommended to set SISPO up in a Windows environment since e.g. the reconstruction algorithms seemed to be more stable. Known differences or problems under Linux will be pointed out. While it should be possible to use a plain Python environment and pip, a miniconda environment manager was used for development. Also a C compiler is necessary. Linux provides the GCC, for Windows it is easiest to install Microsoft Visual Studio with MSVC and MSBuild. Another possibility when using Windows is to use vcpkg. However, previously the openMVG and openMVS ports in vcpkg did not work. Vcpkg can also be used with Linux. However, there were unsolvable problems when using vcpkg so everything was installed natively.

For OpenMVG, OpenMVS and star_cats it is necessary to have the executables in the correct folder for SISPO to function.

Directory structure

`bash sispo ├── build ├── data │   ├── input │   ├── models │   ├── sensor_database │   ├── UCAC4 │   │   ├── u4b │   │   ├── u4i ├── doc │   ├── thesis ├── sispo │   ├── compression │   ├── reconstruction │   └── sim ├── software │   ├── blender │   ├── miniconda │   ├── openMVG │   │   ├── openMVG_build │   │   │   ├── install │   │   ├── openMVG │   ├── openMVS │   │   ├── openMVS_build │   │   │   ├── install │   │   ├── openMVS │   ├── star_cats │   ├── vcpkg `

The directory tree shows the assumed overall folder structure after installation. No sub-folders of the build folder or any files are shown.

To make SISPO perform well, it is beneficial to install the Nvidia CUDA Toolkit in case an Nvidia graphics card is available.

Step-by-step instructions

In the following enumeration, commands intended to be run in a shell are highlighted with a grey box.

  • Clone the GitHub repository onto the local machine git clone https://github.com/YgabrielsY/sispo.git. The project provides a software folder which is intended to be used to install all following software.
  • Setup (conda) environment with dependencies (to software/miniconda folder):
    • orekit 9.3.1, the current version 10.0 had issues when attempted to install. Also orekit needs a data package to function, it is distributed with SISPO in the sim module folder.
    • astropy
    • opencv
    • OpenEXR (For Windows the pre-compiled package found at OpenEXR binaries for Windows needs to be used because the pip or conda version do not work.
    • NumPy
    • Python
  • (Especially Windows) Install vcpkg to software/vcpkg folder, follow instructions at vcpkg
  • Install Blender as a python module (bpy), during development Blender version 2.8 was used.
    • Clone Blender git repository to software/blender/blender git clone git://git.blender.org/blender.git
    • Compile target bpy make bpy, this works also for Windows through the make.bat file provided with Blender
    • If available: Activate CUDA in the cmake project and recompile
    • Install bpy to python environment. Follow these instructions
  • Install OpenMVG, follow instructions at OpenMVG Build or look for hints in the OpenMVG install script in the build folder.
    • Install dependencies according to instructions
    • Clone OpenMVG GitHub repository to software/openMVG/openMVG git clone --recursive https://github.com/openMVG/openMVG.git
    • Build to software/openMVG/build_openMVG folder
    • Install to software/openMVG/build_openMVG/install folder
  • Install OpenMVS, follow instructions at OpenMVS Build or look at the OpenMVS install script in the build folder for hints.
    • Install dependencies according to instructions
    • Clone OpenMVS GitHub repository to software/openMVS/openMVS git clone https://github.com/cdcseacave/openMVS.git
    • Build to software/openMVS/build_openMVS folder
    • Install to software/openMVS/build_openMVS/install folder
  • Install star_cats
    • Clone star_cats GitHub repository to software/star_cats/star_cats git clone https://github.com/Bill-Gray/star_cats.git
    • Build to software/star_cats/build_star_cats make
  • Download UCAC4 star catalog to data/UCAC4, use either:
    • the build/data/download_ucac4.sh script
    • download the folders u4b and u4i directly from e.g. UCAC4