Fertilized Forests Decision Forest library

The fertilized forests library is a general purpose, platform independent, easy to extend, Decision Forest library. It is unmaintained at the moment. For an even more efficient rewrite of the library, see also https://github.com/classner/forpy .

Build status

The construction branch is unstable. Only versions that are completely tested get merged to master, so master is always always stable, with all tests passing and all builds running.

The current status of unstable is as follows:

• Build and tests, core library - Windows
• Build and tests, core library and Python interface - Linux & Mac OS

Compilation & Installation

The project has a CMake build system. The series of

mkdir build
cd build
cmake ..
cmake --build . # add `--config Release` for Windows
cmake --build . --target test  # run the tests (with `--target run_tests` and `--config Release` on Windows).
cmake --build . --target install  # run the installation (... for Windows).

is sufficient to create the project in a Release configuration on all platforms.

The build does require Boost and Eigen to link against. Additionally, it offers to have feature extraction with OpenCV, and to build Python and MATLAB bindings. Hence, the following switches might be interesting for you:

-DCMAKE_BUILD_TYPE={Release|Debug}
-DWITH_PYTHON={OFF|ON}
-DWITH_MATLAB={OFF|ON}
-DWITH_CAFFE={OFF|ON}

-DBOOST_ROOT=...
-DEIGEN3_INCLUDE_DIR=...
-DCAFFE_ROOT_DIR=...
-DCAFFE_CPU_ONLY={OFF|ON}  # Must match the caffe compilation mode.

And there are many more fine-grained options for control. Simply open the file CMakeCache.txt after you ran CMake for the first time. You can change every variable in the file by adding a command line switch -DVARNAME=VALUE to another CMake call to adjust the configuration.

Developing

If you plan to extend or modify the library, you may also need the fertilized-devtools project that provides a code generator to automatically keep the bindings and the serialization code up-to-date.

Troubleshooting

If you want to see how environment setup in general works, see the files appveyor.yaml (Windows) or .travis.yaml (Linux).

If all this fails or you find a bug, please report it on github to let the community know and get it fixed!

Hint: When building with MATLAB on Linux, you might run into an issue with MATLAB not finding GLIBCXX_X.Y.ZZ. It's a known issue of MATLAB and easy to fix. You can find the solution here.

Hint: When building the Python bindings for Python 3 on Linux, you might have to

cd /usr/lib/x86_64-linux-gnu/
sudo ln -s libboost_python-py3[X].so libboost_python3.so

where [X] must be replaced with the appropriate value. This should solve the problem that boost_python3 can not be found.