Find a fragment in the whole image...

why is cv::warpPerspective outputting different values for the same inputs on Python vs. C++?...