Building with cilantro has been tested on Visual Studio 2017 also with CUDA 9.2, but, should in theory, also work with any version of CUDA 10 as well. CUDA and cmake can be installed from their official websites. The community version of Visual Studio is sufficient. OpenCV can be downloaded from the official website as a prebuilt-binaries-and-header package and extracted to a location of your choosing.
This code and all its dependencies MUST be built in 64 bit. In other words, you should select the Win64 (or, for later CMake versions, x64) generator in cmake.
To build cilantro you also need to build its dependence Pangolin. The build-and-installation procedures for Pangolin, cilantro, and GLFW are very similar. Note that cilantro requires the Eigen to be properly installed via CMake, and you might as well use the aready-provided version of Eigen in external/eigen3.4 for this. All of these packages are managed by CMake and you can follow the standard procedure to generate the Visual Studio solution. After solution generation, open it in Visual Studio running as an administrator (right-click the VS icon and choose Run as administrator). You can then build and install the package using the INSALL project inside the solution.
Now you are ready to build this repo. To tell CMake where to find the dependencies, you can set xxx_DIR to the location of xxxConfig.cmake on disk, where xxx can be glfw3_DIR, OpenCV_DIR, and/or any other dependencies. Alternatively, add these as permanent environment variables to simplify future builds. After that, just follow the standard CMake and Visual Studio workflow to build this repo.