How to share cv::Mat for processing between cpp and python using shared memory

Question

I am using shared memory provided by boost/interprocess/ to share the cv::Mat between model and client (both C++). Now I need to use a model in Python. Can you please tell which is the best way to share the cv::Mat between C++ and Python without changing the present client. Thanks.

Answer 1

The task was completed using mapped memory to share the cv::Mat between C++ and Python process.

• C++ - use boost to copy cv::Mat to a mapped shared memory

#include <boost/interprocess/shared_memory_object.hpp>
#include <boost/interprocess/windows_shared_memory.hpp>
#include <boost/interprocess/mapped_region.hpp>
#include <opencv2/opencv.hpp>

using namespace boost::interprocess;

int main(int argc, char* argv[])
{
imgMat= cv::imread("image.jpg");

windows_shared_memory shmem(open_or_create, "shm", read_write, img_size);

mapped_region region(shmem, read_write);

unsigned char* img_ptr = static_cast<unsigned char*> (region.get_address());

std::memcpy(img_ptr , imgMat.data, img_size);

std::system("pause");
}

• Python - read the image from memory using mmap.

import mmap
import cv2 as cv
import numpy as np

if __name__ == '__main__':
  shmem = mmap.mmap(-1, image_size ,"shm")
  shmem.seek(0)
  buf = shmem.read(image_size)
  img = np.frombuffer(buf, dtype=np.uint8).reshape(shape)
  cv.imwrite("img.png", img)
  shmem.close()