laplacian-foveation

Installation

laplace-foveation requires a Python distribution (with NumPy and pybind11 packages installed), OpenCV 4, a C++ compiler and CMake.

Windows

These steps assume the Anaconda Python Distribution is used, OpenCV is downloaded and extracted, and Visual Studio is installed.

• On Windows 11, you might need to install the Media Feature Pack
• Open an Anaconda Prompt and install pybind11 and opencv packages:

  conda install -c conda-forge pybind11
  pip install opencv-python

• Clone this repository and create a build directory in the root directory (laplacian-foveation/build)
• Open CMake and specify the cloned repository directory as source and the created build subdirectory to build the binaries
• The first time you hit Configure some errors will likely be thrown, here are fixes/what you might need to define manually:

  Name	Value
  OpenCV_DIR	%path_to_extracted_opencv_folder%/build
  pybind11_DIR	%path_to_anaconda%/anaconda3/Lib/site-packages/pybind11/share/cmake/pybind11
  PYTHON_EXECUTABLE	%path_to_anaconda%/anaconda3/python.exe

• When Configure is successful, hit Generate and then Open Project to open Visual Studio on the generated files
• In Visual Studio, change Debug/x64 to Release/x64 in the upper bar near the menu and, on the right side, right-click laplacian-foveation and select Build(/Rebuild)
• If the build process succeeded, go to laplacian_foveation/build/Release/ and copy the file pysmooth_foveation.PYD to %path_to_anaconda%/anaconda3/Lib/site-packages/
• Add OpenCV bin directory to PATH (%path_to_extracted_opencv_folder%/build/x64/vc15/bin) -- if you have any Anaconda terminal window open, you need to close and reopen it so it is aware of the change to PATH
• Test the build and installation of laplacian_foveation by running (Python):

  import laplacian_foveation as fv

• The test.py script provides a demo on foveating an image from the COCO dataset:

  conda install -c anaconda scikit-image
  python test.py

Linux

This installation was tested on Ubuntu 22.04.1 LTS version using a Python Virtual Environment. These steps assume that pip is installed and OpenCV is downloaded and extracted.

• Create and activate a Python Virtual Environment.

  python3 -m venv my_env_project
  source my_env_project/bin/activate

• With the Python Virtual Environment activated, install pybind11, NumPy and opencv packages:

  pip install pybind11
  pip install numpy
  pip install opencv-python

• Clone this repository and create a build directory in the root directory (laplacian-foveation/build):

  cd %path_to_clone_directory%/laplacian_foveation
  mkdir build

• Open CMake and specify the cloned repository directory as source and the created build subdirectory to build the binaries
• The first time you hit Configure some errors will likely be thrown, here are fixes/what you might need to define manually:

  Name	Value
  OpenCV_DIR	%path_to_extracted_opencv_folder%/build
  pybind11_DIR	%path_to_virtualenv%/lib/site-packages/pybind11/share/cmake/pybind11

• When Configure is successful, hit Generate and then build/compile the project (Unix Makefiles) in the terminal:

  cd build
  make

• If make fails, edit the file laplacian_foveation/src/python_bindings/conversions.cpp replacing #include <numpy/ndarrayobject.h> with #include "%path_to_virtualenv%/lib/python3.x/site-packages/numpy/core/include/numpy/ndarrayobject.h", where python3.x must be the Python3 version installed in your Python Virtual Environment.
• If the build/compile process succeeded, go to laplacian_foveation/build/ and copy the file laplacian_foveation.cpython-310-x86_64-linux-gnu.so to %path_to_virtualenv%/lib/site-packages/
• Test the build and installation of laplacian_foveation by running (Python):

  import laplacian_foveation as fv

• The test.py script provides a demo on foveating an image from the COCO dataset:

  pip install scikit-image
  python test.py