Remote Modules

What is a Remote Module

Remote modules allow anyone to extend the functionalities of OTB without being part of the core project repository. They can have different licenses than the main OTB repository. Those modules are just like regular modules, except they are not distributed inside OTB source code. Under some conditions (dependencies, official acceptance process, etc.), we are also able to distribute your remote module in the official standalone binaries.

List of available modules

Official OTB modules

“Official” remote modules are just regular modules that we consider particularly useful for users. Some of those are shipped in the binary packages so you don’t need to build them yourself.

  • DiapOTB : Potential events analyze by highlighting differences between SAR images

    • Repository: https://gitlab.orfeo-toolbox.org/remote_modules/diapotb
    • License: Apache License 2.0
    • Description: The differential SAR interferometry (DInSAR) technique relies on the processing of two SAR images of the same portion of the Earth’s surface taken at different time. The aim is to analyze potential events (earthquake, destruction, …) by highlighting differences between SAR images. DInSAR involves a set of tools such as creation of deformation grids , coregistration or building of interferograms. The Orfeo Toolbox remote module DiapOTB contains all necessary steps and allows to launch a complete DInSAR chain. This module has been used with Sentinel-1 data with satisfactory results. This module is a port from DIAPASON tool, integrated in the ESA GeoHazards TEP.
  • otb-mosaic : Image mosaiking

    • Repository : https://github.com/remicres/otb-mosaic
    • Author: Rémi Cresson
    • License: CeCILL-B
    • Description: This module provides one application dedicated to images mosaicking with several compositing methods available. This module is now part of OTB since version 7.0.0
  • otb-bv : Estimation of biophysical variables

    • Repository: https://gitlab.orfeo-toolbox.org/jinglada/otb-bv
    • Author: Jordi Inglada
    • License: GNU Affero General Public License v3.0
    • Description: The otb-bv project allows for estimation of biophysical variables (LAI, fAPAR, fCover) from remote sensing images using machine learning nonlinear regressions for the inversion of PROSPECT+SAIL model.
  • Phenotb : Extract phenological information from time profiles

    • Repository: https://gitlab.orfeo-toolbox.org/jinglada/phenotb/
    • Author: Jordi Inglada
    • License: GNU Affero General Public License v3.0
    • Description: This module implements several algorithms allowing to extract phenological information from time profiles. These time profiles should represent vegetation status as for instance NDVI, LAI, etc.
  • otbFFSforGMM : Large-scale feature selection with Gaussian mixture models

  • GRM : Generic region merging segmentation

    • Repository: http://tully.ups-tlse.fr/lassallep/grm/tree/master
    • Author: Pierre Lassalle
    • License: GPL v3
    • Description: This module provides the GRM OTB application to perform multi-scale region-merging segmentation on satellite images. Three local homogeneity criteria are available: the Baatz & Schäpe criterion, the Full Lambda Schedule criterion and the simple Euclidean Distance criterion. This module was contributed by Pierre Lassalle who also provides a tutorial to learn how to use the library.
  • SertitObject : Object oriented image analysis

    • Repository: https://github.com/sertit/SertitObject
    • Author: SERTIT - University of Strasbourg
    • License: CeCILL-B
    • Description: This module provides two OTB applications dedicated to object-oriented image analysis.
  • Temporal gap-filling : perform temporal gap-filling in image time series

    • Repository: http://tully.ups-tlse.fr/jordi/temporalgapfilling.git
    • Author: Jordi Inglada
    • License: GNU Affero General Public License v3.0
    • Description: This module provides classes and one application to perform for temporal gap-filling in image time series (linear and spline interpolators are provided).

Community modules

  • Feature selection :

    • Repository: https://github.com/boussaffawalid/FeatureSelection
    • Author: Walid Boussafa and Nesrine Chehata
    • License: All rights reserved (without license granting more rights, copyright fully applies, this component can not be used without the explicit and prior authorization of the copyright owner).
    • Description: This module contains a FeatureSelection application based on the FST3Lib.
  • OTBTensorflow (otbtf) : generic, multi purpose deep learning framework, targeting remote sensing images processing

    • Repository: https://gitlab.irstea.fr/remi.cresson/otbtf
    • Author: Rémi Cresson
    • License: Apache License 2.0
    • Description: This remote module of the Orfeo ToolBox provides a generic, multi purpose deep learning framework, targeting remote sensing images processing. It contains a set of new process objects that internally invoke Tensorflow, and a bunch of user-oriented applications to perform deep learning with real-world remote sensing images. Applications can be used to build OTB pipelines from Python or C++ APIs.

Installation and usage

Build possibilities

Your remote module can be built inside the OTB source tree or outside as an external CMake project with an existing OTB installation.

  • Building against a build tree

    In this case you have compiled OTB from source , the cmake configuration will be done inside OTB build directory.

    Note that there are two ways of compiling:

    • Build as a module inside OTB, in which case build files will be written to the OTB build tree as other modules. Main benefit is that this will enrich the current OTB build with your new module, but you need to have write access to the build directory. For this type of build, the cmake configuration is easy, see below the compilation chapter
    • Build as a standalone CMake project, in which case build files will remain in the module build folder. This build is fully independent from the build (or install) directory, but the module will not be recognized as an OTB module (still you will be able to use its binaries and libraries).

    This behaviour is controlled by the cmake option OTB_BUILD_MODULE_AS_STANDALONE, which is OFF by default (hence first behaviour). There are other cmake options to set in order to build it as standalone, this is described below.

  • Building against an installed OTB

    In this case, only the second behaviour (build as standalone) is available. This requires to specify cmake options for the build :
    • Set the module to build as standalone with OTB_BUILD_MODULE_AS_STANDALONE=ON
    • Set the following option -DCMAKE_CXX_FLAGS=-D_GLIBCXX_USE_CXX11_ABI=0 (see bellow)
    • Set the OTB install directory with the variable OTB_DIR
    • Set an install folder for your libraries CMAKE_INSTALL_PREFIX=/theModulePath/install
    • Set the runtime path RPATH of your libraries to your install/lib folder DCMAKE_INSTALL_RPATH=/theModulePath/install/lib
    • Tell cmake to set runtime path using link path : CMAKE_INSTALL_RPATH_USE_LINK_PATH=TRUE (permits to avoid modifying your LD_LIBRARY_PATH)

The OTB binaries are compiled with the option GLIBCXX_USE_CXX11_ABI set to 0. This is why you have to build your remote module with the same option. If you don’t, you will encounter a “symbol not found” error message while running your application.

Compilation and Installation

  • If you choose the inside OTB build, your module will be built with the rest of the OTB project. To add the module to the compilation process you have two options:

    • use OTB to retrieve automatically the official remote module (not applicable for community/self remote modules) that you want to build. All you have to do is to call cmake configuration in OTB build dir to activate Module_TheModuleName
    • clone the module yourself (necessary if you use a community module, or your own) and copy the folder to OTBSource/Modules/Remote, this will trigger a new option in CMake config named Module_TheModuleName which is OFF by default.

    Open a terminal and run :

cd /PathToOTB/build
cmake -DModule_TheModuleName=ON
make install

the applications of your module will be installed in the same folder as OTB applications

  • If you choose the OTB install build :
mkdir /Path/to/Module/build && cd /Path/to/Module/build
cmake -DOTB_DIR=/PathTo/OTB/install -DOTB_BUILD_MODULE_AS_STANDALONE=ON
-DCMAKE_CXX_FLAGS=-D_GLIBCXX_USE_CXX11_ABI=0
-DCMAKE_INSTALL_PREFIX=/theModulePath/install -DCMAKE_INSTALL_RPATH=/theModulePath/install/lib
-DCMAKE_INSTALL_RPATH_USE_LINK_PATH=TRUE ../
make install

the applications will be installed in /theModuleInstallFolder/lib and the binaries will be available in /theModuleInstallFolder/bin

Usage

  • For an inside OTB build, verify that you have the OTB environment variables set, and call your application :
otbcli_MyModuleApp arg1 ... argX
  • For an OTB install build, you have to add your /theModuleInstallFolder/lib to the variable OTB_APPLICATION_PATH, and theModuleInstallFolder/bin to the PATH :
export OTB_APPLICATION_PATH=/theModuleInstallFolder/lib:$OTB_APPLICATION_PATH
export PATH=/theModuleInstallFolder/bin:$PATH

We strongly recommend adding these exports in your .bashrc in order to make your applications available system wise

Writing your own remote module

This section will guide you step by step to create your own remote module. To start, you can fork our remote module template project: Remote Module Template. Each module is made of different components, described in the following sections.

The otb-module.cmake file

This file is mandatory. It follows the CMake syntax, and has two purposes:

  • Declare dependencies to other modules,
  • Provide a short description of the module purpose.

These purposes are fulfilled by a single CMake Macro call:

otb_module(TheModuleName DEPENDS OTBModule1 OTBModule2 ... OTBModuleN DESCRIPTION "A description string")

Note: You can use the keyword TESTDEPENDS to declare module dependencies that only applies to the tests.

The CMakeLists.txt file

The CMakeLists.txt file is mandatory. It contains only a few things. First, it declares a new CMake project with the name of the module:

project(OTBTheModuleName)

Second, if the module contain a library (see src folder section below), it initializes the TheModuleNameLIBRARIES CMake variable (if your module only contains headers or template code, skip this line):

set(OTBTheModuleName_LIBRARIES OTBTheModuleName)

You can build your remote modules inside the OTB source tree by copying your source inside the directory Module/Remote or against an existing OTB build tree (note that it does not work with an install version of OTB).

The configuration below will handle both cases and take care of all the CMake plumbing of the module:

if(NOT OTB_SOURCE_DIR)
  find_package(Boost REQUIRED COMPONENTS filesystem serialization)
  find_package(OTB REQUIRED)
  list(APPEND CMAKE_MODULE_PATH ${OTB_CMAKE_DIR})
  # The Python interpreter is needed for Python tests
  set(Python_ADDITIONAL_VERSIONS "3")
  find_package( PythonInterp REQUIRED)
  include(UseOTB)
  include(OTBModuleExternal)
else()
  otb_module_impl()
endif()

The overall file should look like this:

cmake_minimum_required(VERSION 3.10.0)
project(OTBTheModuleName)
set(OTBTheModuleName_LIBRARIES OTBTheModuleName)

if(NOT OTB_SOURCE_DIR)
  find_package(Boost REQUIRED COMPONENTS filesystem serialization)
  find_package(OTB REQUIRED)
  list(APPEND CMAKE_MODULE_PATH ${OTB_CMAKE_DIR})
  # The Python interpreter is needed for Python tests
  set(Python_ADDITIONAL_VERSIONS "3")
  find_package( PythonInterp REQUIRED)
  include(UseOTB)
  include(OTBModuleExternal)
else()
  otb_module_impl()
endif()

Remarque: the command find_package(Boost) is called before
find_package(OTB). This is due to the fact that FindBoost.cmake is
integrated to cmake, so it is better to use the official command
rather than the one integrated to the OTB.

The include folder

The include folder will contain all your headers (*.h files) and template method files (*.hxx or *.hxx). It does not require any additional file (in particular, no CMakeLists.txt file is required).

The src folder

The src folder contains the internal implementation of your module:

  • It typically contains cxx source files that will be compiled into a library.
  • It can contain header files for classes used only within the implementation files of your module. Any header file present in the src folder will not be installed, and will not be available to other modules depending on your module.

If your modules is made of template only code, you do not need a src folder at all.

If present, the src folder requires a CMakeLists.txt file.

The first part of the CMakeLists.txt file is classical, as it builds the library and links it:

set(OTBTheModuleName_SRC
    sourceFile1.cxx
    sourceFile2.cxx
    sourceFile3.cxx
    ...
    sourceFileN.cxx)

add_library(OTBTheModuleName ${OTBTheModuleName_SRC})

target_link_libraries(OTBTheModuleName ${OTBModule1_LIBRARIES} ${OTBModule2_LIBRARIES} ... ${OTBModuleN_LIBRARIES})

Notes:

  • Library name should match the one declared in the root CMakeLists.txt when setting CMake variable TheModuleName_LIBRARIES,
  • Linked libraries should match the dependencies of your module declared in the root otb-module.cmake file.

The last line of CMake code takes care of installation instructions:

otb_module_target(OTBTheModuleName)

The overall CMakeLists.txt file should look like:

set(OTBTheModuleName_SRC
    sourceFile1.cxx
    sourceFile2.cxx
    sourceFile3.cxx
    ...
    sourceFileN.cxx)

add_library(OTBTheModuleName ${OTBTheModuleName_SRC})

target_link_libraries(OTBTheModuleName ${OTBModule1_LIBRARIES} ${OTBModule2_LIBRARIES} ... ${OTBModuleN_LIBRARIES})

otb_module_target(OTBTheModuleName)

The app folder

The app folder contains the code of applications shipped with your module. If your module has no application, you do not need the app folder.

Notes: If your module contains application (and an app folder), do not forget to add the ApplicationEngine in the dependencies listed in the otb-module.cmake file.

In addition to the applications source code, the app folder should contain a CMakeLists.txt file as follows.

For each application, a single call otbcreateapplication is required:

otb_create_application(
  NAME           TheModuleApplication1
  SOURCES        TheModuleApplication1.cxx
  LINK_LIBRARIES ${OTBModule1_LIBRARIES} ${OTBModule2_LIBRARIES} ... ${OTBModuleN_LIBRARIES})

The test folder

This folder contains tests of the module. If your module has no test in it (which is not recommended, you do not need it).

The test folder should contain the source files of tests, as well as a CMakeLists.txt file. This file will contain the following.

First, indicate that this folder contains tests.

otb_module_test()

Then, build the test driver:

set(OTBTheModuleNameTests
    testFile1.cxx
    testFile2.cxx
    ...
    testFileN.cxx)

add_executable(otbTheModuleNameTestDriver ${OTBTheModuleNameTests})

target_link_libraries(otbTheModuleNameTestDriver ${OTBTheModuleName-Test_LIBRARIES})

otb_module_target_label(otbTheModuleNameTestDriver)

Finally, you can add your tests:

otb_add_test(NAME nameOfTheTest COMMAND otbTheModuleNameTestDriver
             --compare-image ${EPSILON_8} ... # baseline comparison if needed
             nameOfTheTestFunction
             testParameters)

If your module contains one or more application in the app folder, you should also write tests for them, in the test folder. Running an application test is easily done with the helper macro otbtestapplication:

otb_test_application(NAME   nameofApplication1Test1
                      APP  TheModuleApplication1
                      OPTIONS -in1 ${INPUTDATA}/input1.tif
                              -in2 ${INPUTDATA}/input2.tif
                              -out ${TEMP}/nameofApplication1Test1_result.tif
                      VALID   --compare-image ${EPSILON_8}
                              ${BASELINE}/nameofApplication1Test1_result.tif
                              ${TEMP}/nameofApplication1Test1_result.tif)

To add a test executed by a Python script using OTB Applications bindings:

set(TEST_DRIVER otbTestDriver
    --add-before-env OTB_APPLICATION_PATH $<TARGET_FILE_DIR:otbapp_EmptyApp> )

otb_add_test(NAME otbEmptyScriptTest
  COMMAND ${TEST_DRIVER} Execute ${PYTHON_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/EmptyScript.py)

Overall CMakeLists.txt should look like:

otb_module_test()

set(OTBTheModuleNameTests
    testFile1.cxx
    testFile2.cxx
    ...
    testFileN.cxx)

add_executable(otbTheModuleNameTestDriver ${OTBTheModuleNameTests})

target_link_libraries(otbTheModuleNameTestDriver ${OTBTheModuleName-Test_LIBRARIES})

otb_module_target_label(otbTheModuleNameTestDriver)

otb_add_test(NAME nameOfTheTest COMMAND otbTheModuleNameTestDriver
             --compare-image ${EPSILON_8} ... # baseline comparison if needed
             nameOfTheTestFunction
             testParameters)

otb_test_application(NAME otbEmptyAppTest
                 APP  EmptyApp
                 )

Use Python OTB & GDAL dependency in your module

If your module have a Python part, which is using OTB python bindings, you should encounter some troubles with the binary version, here is how to fix it:

First install OTB on your platform. See the related documentation to install OTB on your system..

Then, you’ll need a version of GDAL which is compatible with your OTB version.

  • In case you’re using OTB binary distribution, you’ll need to patch the files provided.

    • For that purpose you can find this simplified and generic version of gdal-config in the OTB source tree : Documentation/CookBook/Scripts/gdal-config. Just drop it into the bin/ directory where you’ve extracted OTB. This will permit pip install gdal==vernum to work correctly.

    • You’ll also have to patch otbenv.profile to insert OTB lib/ directory at the start of $LD_LIBRARY_PATH. This will permit python3 -c 'from osgeo import gdal' to work correctly.

      # For instance, type this, once!
      echo 'LD_LIBRARY_PATH="${CMAKE_PREFIX_PATH}${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"' >> otbenv.profile
      

In case you’ve compiled OTB from sources, you shouldn’t have this kind of troubles.

Sharing your module

To make your remote module available to others, you should first host the module code on a publicly available git repository. If you do not have access to a git server, bitbucket or github can provide this service for you. Then you should provide a CMake file named TheModuleName.remote.cmake for inclusion in the Modules/Remote folder in OTB source tree.

This file should contain the following:

# Contact: Author name <author email address>

otb_fetch_module(TheModuleName
  "A description of the module, to appear during CMake configuration step"

  GIT_REPOSITORY http_link_to_a_git_repository_hosting_the_module
  GIT TAG the git revision_to_checkout
)

This file should be provided along with your remote module inclusion proposal email to the otb community list : acceptance of remote module is submitted to vote on otb-developers. If accepted, your CMake file will be placed into the Modules/Remote folder inside OTB source tree.

Important Note: a remote module bringing in new third party dependencies will not be included in binary packages.

During the OTB release process, all modules complying with the remote module release policy will be packaged along with standard modules.

A remote module can be removed from Modules/Remote (this only requires to remove the CMake file describing it), if:

  • It does no longer comply with the remote module acceptance policy (in which case the decision is submitted to vote on OTB’s forum),
  • The author of the remote module ask to remove it.
  • Remote module acceptance policy

So as to get your module accepted as an official remote module, you should comply with the following:

  • Remote module source code should be hosted on a publicly available Git repository
  • Author of the remote module should be identified and registered to OTB’s forum
  • Author of the remote module accepts to be contacted by developers or users regarding issues with his module (on a best effort basis),
  • Remote module source code should comply with OTB style as much as possible,
  • Remote module source code should be documented using doxygen tags,
  • Remote module should provide a minimal set of tests to ensure build of template code and basic non-regression testing,
  • Remote module should come with a form of documentation (website, publication, readme file …)
  • Remote module should not embed code from any third party software (unless strong arguments are given by the author, in which case an exception can be made),
  • Remote module should avoid depending on new third parties if possible,
  • Remote module author should be the copyright owner and comply with licences of any third party, which in turn should comply with terms of OTB licence (to be reviewed by PSC)
  • Author of remote module should provide a small description of the remote module to be added on OTB website.

An internal module should never depend on a remote module whatsoever.

  • Remote module release policy

During the OTB release process, a remote module will be included in source and binary packages if dashboard submission exist and show that the remote module:

  • Builds on all plateform
  • Passes all tests on the reference platform
  • Does not have any test crashing (i.e. failing with core dump or memory issues) on remaining platform
  • The remote module complies with the remote module acceptance policy at the time of the Release Candidate

Developers will notify remote modules authors of existing issues at Release Candidate. If by 3 day to the final release dates, some issues listed above still exist, the remote module will be removed from the release source and binary packages.

Use Continuous Integration

We encourage you to use a CI platform during the development of your module.

In the remote module template, we provide two files which permits to put your module on a CI
  • ci.cmake => used to build your module
  • travis.yml => used by Travis-CI to call cmake using ci.cmake script and launch your tests

Those files enable your module to use Travis-CI, which is the github continuous integration platform. if your repository is on Gitlab, you can make a mirror of your repo to github

To modify the travis.yml: use the manual

You have two options to use OTB in your CI :

  • Get an installer from the website and install it in the install: part of travis.yml. You will have to build your module as standalone
env:
  global:
    - OTB_URL=https://www.orfeo-toolbox.org/packages/archives/OTB
    - OTB_VER=7.2.0
    - OTB_OS=Linux64
    - OTB_PKG_EXT=run

install:
  - export OTB_PKG="OTB-${OTB_VER}-${OTB_OS}.${OTB_PKG_EXT}"
  - wget ${OTB_URL}/${OTB_PKG}
  - chmod +x ${OTB_PKG}
  - ./${OTB_PKG} --target xdk

script:
  - command to build and test your module here
  • Build a docker image containing an otb build tree. Run this docker image in travis and build your module against this build tree
before_script:
  - docker pull YourOTBImage

script:
  - docker run -it YourOTBImage /bin/bash -c "ctest -VV -S ci.cmake"

In the before script section, you have to set the environment variables. For this you can create a script called activate_env.sh in your main module folder which contains:

export OTB_RESULT_DIR=/home/travis/build/yourName/yourModule/data/OutputTest
export OTB_DATA_DIR=/home/travis/build/yourName/yourModule/data
export PYTHONPATH=/OTB_InstallDir/lib/otb/python
export OTB_APPLICATION_PATH=/home/travis/build/yourName/yourModule/install/lib:/OTB_InstallDir/lib/otb/applications

and call it in travis.yml:

before-script:
   - source activate_env.sh

You can test your module with one line in the travis.yml:

script:
   - ctest -VV -S ci.cmake

This command builds your project and launches the tests.