This is the repository of my project for the CMU course 16-823 Physics-Based Methods in Vision. The collected data and generated figures are available at this link.

Some scripts included in this repo:

• capture_frames.py provides the live feed of the camera. Collect the current frame by pressing c.
• plot_histograms.py contains all the code we used to generate figures in the report.

Some guides on compiling the Royale SDK (downloadable from the official website of PMD):

Installation on Windows

The binary files in roypy/ are for Python 3.10 on Windows. If you want to build on your version of Python:

First, install Visual Studio, CMake, and swig. Make sure they are in your PATH or install them with scoop / winget:

winget install cmake
scoop install swig

Then, activate your virtual environment, in which you should install numpy, matplotlib, and pywin32. To make sure it compiles windows, define ssize_t in swig/roypy.h by

#if defined(_MSC_VER)
    #include <BaseTsd.h>
    typedef SSIZE_T ssize_t;
#endif

or if that doesn't work, try:

#if !defined(ssize_t) && !defined(__MINGW32__)
    #if defined(_WIN64)
        typedef __int64 ssize_t; 
    #else
        typedef long ssize_t;
    #endif
#endif

With the correct python path set, run

cmake -S .\swig -B .\_build -G "Visual Studio 16 2019" -Droyale_USE_NUMPY_IN_ROYPY=on

You may also include -T ClangCL if you prefer. You mays also see if other versions of Visual Studio work, and as far as I tested, VS2019 should work well with Python 3.10 and Python 3.11.

Build and copy the built files to your project path, amke sure there is a file called _roypy.pyd:

cmake --build .\_build\ --config Release
cp .\_build\bin\* .\

you might be seeing some warnings, but it's OK to ignore them. Now you may try if everything works by

python .\sample_opencv.py

Some issues on Linux

• The binary file you need now is _roypy.so, and it may not be found in _build/bin; check if it is in _build/.

• If you see some weird "QObject" and "QThread" errors, install opencv from source:

  sudo apt install libgtk2.0-dev pkg-config # this is necessary for imshow()
  pip install --no-binary opencv-python opencv-python

  Run pip cache purge if necessary.