Oh you’re right about that. In this case, I’m well versed in these topics and algorithms, having done a lot of it manually before.
OpenCV is painful to work with in some regards. Specifically the 3D calibration and pose estimation portions. I haven’t used the entire library but the portions I have used vary a lot. Some parts are amazing. Others are really painful.
I think there’s a better ways of implementing the image processing related stuff.
Their library grows and updates very quickly. The portion on machine learning, neural nets, and AI are where they focus. I don’t see updates or changes being made to the image processing portions of the library. In my opinion, they need QOL updates along with new functions that are simple but missing.
Some types like cv::Mat are so overloaded the types are really weakly enforced. cv::Mat will accept almost any type. You can enforce some types using things like cv::Mat_< float> but that breaks some things like matrix multiplication. There’s a lot of type casting required.
One of worst things about the library is the typing but I feel that way about c++ in general
Edit: Let me clarify. I don’t want to rewrite the whole thing. Just specific portions, like the image processing basics, detecting some stuff in images, and some photogrammetry portions