I’m actually trying to run a script containing opencv function on cuda.
More precisely, I am trying to parallelize the undistort function.
On python, I’ve tried to use jit through numba, which didn’t get me anywhere since opencv and numba do not seem to be compatible.
On C++, I’ve tried to parallelize, following the following example (https://developer.nvidia.com/blog/even-easier-introduction-cuda/)
But it do not work because the undistort function seems to be a host function and not a device one.
Moreover, undistort is not part of opencv already accelerated function like in cv::gpu.
Is there a way to accelerate it using opencv and cuda libraries and modules ?
Or should I create my own accelerated version of undistort() ? (knowing that it is not the only function I’ll have to accelerate)