This is a minimal example to build a Python extension that uses both numpy and opencv. In some applications, postprocessing can be cpu-intensive, e.g. drawing contours for image segmentation moidels. These logics can be implemented in C++ to better utilize multi-core.
This example demonstrates:
- How to interface between Python and C++
- How to convert a numpy array to cv::Mat with zero copy
- How to write a setup.py to set proper include directories and link libraries.
- How to call opencv/numpy functions in C++, parse Python variables and construct returning variables.
Use conda to manage dependencies, make sure you run the setup.py inside an active conda environment. You also need to install numpy and opencv, e.g. conda install -c conda-forge opencv
python setup.py build_ext --inplace
python test_example.py
