RTOC

Real Time Object Classification

Brief:

This project is derived from the master thesis by Lisa O. Aagensen, DTU Physics. The initial goal was to not only speed up the process in acquiring images and processing these images for data extraction, but also to work towards a real-time approach where images are acquired and analyzed while found objects are classified by a loaded ML model.

Video on how to use the application

Build instructions

Note: for easy configuration, debugging etc., it can be highly recommended use QtCreator (bundled with most Qt installations) when working with the project.

Dependencies:

Boost (1.66.0 or higher) https://www.boost.org/
OpenCV (3.4.1 or higher) https://opencv.org/
Qt (5.8 or higher) https://www.qt.io/download
CMake (3.8 or higher) https://cmake.org/
[For IronMan framegrabber] Silicon Software Runtime 5.5.1 https://silicon.software/blog-runtime-551/

Linux

MacOS

Note: As we have encountered some bugs, macOS 10.13 High Sierra is not recommended.

Note: This build only includes acquisition from folder as for now.

This build have both been tested with default gcc and clang compilers from Xcode Command Line Tools on macOS 10.12 Sierra, thus this setup is recommended.

To install Xcode Command Line Tools, run

$ xcode-select --install

Afterwards, you can test if the tools are successfully installed by running

$ xcode-select -p
/path/to/CommandLineTools

Regarding the dependencies, use your favorite package manager. We recommend brew, as then installation should be rather smooth, simply run

$ brew install boost
$ brew install opencv
$ brew install qt

To get CMake, it is highly recommended to get the GUI version. Latest binaries are found here.

Once your working environment is setup, you should be able to start building:

Launch the CMake app
Select the path to your source and build directories.
Configure
- If warnings are created from missing directories, set the right paths and reconfigure.
Generate

Now you are fully ready to initialize the build

Run $ cd /path/to/build
Run $ make

Windows

On Windows, a compiler needs to be available - make sure that your compiler version matches the one which your boost version has been built with. Microsoft Visual Studio Community can be downloaded for free, to access the MSVC compiler.

Load the CMakeLists.txt file in the CMake gui and run configuration. If all dependencies are available and all required paths are set, the CMake configuration should report success. In the project directory, run

cmake.exe --build . --target all

to build the app.

A new(er) version of boost is currently required for building RTOC. On Windows,we explicitely define the boost .lib files which are linked to. Set environment variable BOOST_ROOT to the path of your compiled boost libraries and make sure that the name of your libraries is equal to the libboost_...1##.lib files which are specified in the CMakeLists.txt file. Currently, we build RTOC with Boost 1.66 using the Visual Studio 2017 compiler (vc141). If you are using another boost version, or your boost libraries named otherwise, you must manually configure the CMakeLists.txt file to suit your needs.

Building with SiliconSoftware support

RTOC contains an interface to a SiliconSoftware microEnable 5 VQ8-CXP6D ironman framegrabber card. To build support for this interface, a SiliconSoftware runtime installation must be present on the system. RTOC has been built using SiliconSoftware Runtime v.5.1.1.

To enable support for the framegrabber, one must set the "BUILD_IRONMAN" flag in the CMake configuration to "ON". The configuration of this can be a bit tricky, and sometimes fails - running the CMake configuration a couple of times in succession usually fixes the issues (These are issues in the CMake files provided by the SiliconSoftware libraries.)

mortbopet/RTOC