MES Tools to Train and Test Models

MES Tools to Train and Test Models

Overview

These are command-line executables for training, testing, and manipulating models, as well as other utilities developed by MES.

You can download a prebuilt release from the Release page. The release is built with glibc 2.17 and will work out of the box on systems with glibc ≥ 2.17 (e.g., Ubuntu >= 13.04, Centos>=7, eg.). If you need to build from source, follow the instructions below.

For general instructions on how to use prebuilt binaries, refer to the Prebuilt Releases section in the main Installation Guide.

Installation Steps

1. Install Boost Libraries: Follow the instructions in the Install Boost Libraries section of the main Installation Guide. Boost is a core dependency for MES Tools. If you compile Boost from source, ensure BOOST_ROOT is set correctly.
2. Clone the required repositories:

   git clone git@github.com:Medial-EarlySign/MR_Tools.git
   git clone git@github.com:Medial-EarlySign/medpython.git MR_LIBS
   

3. Change to the MR_Tools directory:

   cd MR_Tools
   

4. Configure Build Paths: Instead of directly editing CMakeLists.txt, it's recommended to pass configuration variables via CMake arguments or environment variables.

   • MR_LIBS Path: The build system needs to know the location of your medpython (MR_LIBS) clone. If your directory structure is:

     Root Directory
     ├── MR_LIBS
     └── MR_Tools
     

     then the build script should automatically detect MR_LIBS. Otherwise, you might need to set the LIBS_PATH environment variable before running the build script, or pass it as a CMake argument.

   • BOOST_ROOT Path (if compiled from source): If you compiled Boost from source, ensure the BOOST_ROOT environment variable is set to your Boost installation directory (e.g., /home/user/my_builds/Boost). The build system will use this to find Boost.

   Example of setting environment variables before building:

   export LIBS_PATH="/path/to/your/medpython/clone" # e.g., /home/user/Documents/MES/MR_LIBS
   export BOOST_ROOT="/path/to/your/compiled/Boost" # Only if you compiled Boost from source
   

   5. Build the tools:

   AllTools/full_build.sh
   

A full docker image for compilation steps can be found under this link:

• 01.basic_boost A base docker image with Boost
• 02.build_tools A build with tools prepared for usage

Verification

To verify that the MES Tools have been successfully built, you can try running one of the command-line executables.

# After sourcing your environment setup script (e.g., source ~/setup_mes_env.sh)
Flow --version

This command should output the version information for the Flow executable, indicating a successful build and correct environment setup.

Common Issues

1. Can't find Boost libray errors in compilation - Please delete the "./build" folder to recreate all Makefiles again. It holds some bad settings of Boost in cache.

2. Can't find boost_atomic.so in runtime: This usually indicates a linking issue. Ensure that your BOOST_ROOT (if compiled from source) is correctly set and that LD_LIBRARY_PATH includes the Boost library directory. The full_build.sh script and CMakeLists.txt should ideally handle linking boost_atomic automatically if it's a required component. If not, you might need to explicitly add it to the BOOST_LIBS in the CMakeLists.txt.

3. Running an executable may fail with:

   $> Flow ...
   Flow: error while loading shared libraries: libboost_regex.so.1.85.0: cannot open shared object file: No such file or directory
   

   This indicates the Boost shared libraries are not found at runtime. Set LD_LIBRARY_PATH to point to your Boost lib directory. Add to your ~/.bashrc or run once per session:

   export LD_LIBRARY_PATH=/path/to/BOOST_ROOT/lib
   

   If Boost was installed via the system packages (Ubuntu 22.04 uses 1.74, 24.04 uses 1.83), install the appropriate dev packages instead of adjusting LD_LIBRARY_PATH:

   sudo apt install libboost-program-options1.83-dev -y
   

   Replace 1.83 with 1.74 on Ubuntu 22.04 when needed.