on some CUDA versions creating/destroying cufftPlan2d is very time consuming
we now create them in ConvolveImpl::create() and destroy them in the dtor
this solves issue #3385
Fixes two errors when building with the options WITH_CUDA=ON and BUILD_CUDA_STUBS=ON on a machine without CUDA.
In the cudaarithm module, make sure cuda_runtime.h only gets included when CUDA is installed.
In the stitching module, don't assume that cuda is present just because cudaarithm and cudawarping are present (as is the case when building with the above options).
original commit: 22ee5c0c4d
Add python bindings to cudaobjdetect, cudawarping and cudaarithm
* Overload cudawarping functions to generate correct python bindings.
Add python wrapper to convolution funciton.
* Added shift and hog.
* Moved cuda python tests to this repo and added python bindings to SURF.
* Fix SURF documentation and allow meanshiftsegmention to create GpuMat internaly if not passed for python bindings consistency.
* Add correct cuda SURF test case.
* Fix python mog and mog2 python bindings, add tests and correct cudawarping documentation.
* Updated KeyPoints in cuda::ORB::Convert python wrapper to be an output argument.
* Add changes suggested by alalek
* Added changes suggested by asmorkalov
