NVIDIA_OPTICAL_FLOW_2_0_INTEGRATION

[vchiluka5]

This PR integrates NVIDIA'S Optical Flow SDK 2.0.

• [*] I agree to contribute to the project under Apache 2 License.
• [*] To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
• [*] The PR is proposed to proper branch
• [*] There is reference to original bug report and related work
• [*] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• [*] The feature is well documented and sample code can be built with the project CMake

force_builders=Custom
buildworker:Custom=linux-4
build_image:Custom=ubuntu-cuda:18.04

[vchiluka5]

I see python bindings error. @alalek Can you please suggest what i can change to fix those?

[alalek]

You can disable bindings generation for this class:

• remove _W from CV_EXPORTS_W
• remove CV_WRAP

[vchiluka5]

You can disable bindings generation for this class:

• remove _W from CV_EXPORTS_W
• remove CV_WRAP

we will need python bindings since some of our users use python version of above class.
is there any other way to solve this error?

[alalek]

we will need python bindings

Consider adding simple test for Python API (don't check accuracy, only parameters passing/return types/sizes).
Use this as an example: https://github.com/opencv/opencv_contrib/blob/4.5.1/modules/cudaobjdetect/misc/python/test/test_cudaobjdetect.py

[alalek]

Thank you for update!

[alalek]

Well done!

[tomoaki0705]

Hi @vchiluka5
Since this PR was merged, I see build failure on Jetson TX1 and TX2.
The build passes on Xavier, NX and Nano, which has CUDA 10.0 or later, but my TX1 and TX2 still has CUDA 8.0.

The error message is as following.

[ 57%] Building NVCC (Device) object modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/cuda_compile_1_generated_nvidiaOpticalFlow.cu.o
/opencv_contrib/modules/cudaoptflow/src/cuda/nvidiaOpticalFlow.cu(76): error: no instance of overloaded function "surf2Dwrite" matches the argument list
            argument types are: (short2, cudaSurfaceObject_t, unsigned long, int, cudaSurfaceBoundaryMode)

1 error detected in the compilation of "/tmp/tmpxft_00001b35_00000000-7_nvidiaOpticalFlow.cpp1.ii".
CMake Error at cuda_compile_1_generated_nvidiaOpticalFlow.cu.o.Release.cmake:279 (message):
  Error generating file
  /opencv/build/modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/./cuda_compile_1_generated_nvidiaOpticalFlow.cu.o


modules/cudaoptflow/CMakeFiles/opencv_cudaoptflow.dir/build.make:70: recipe for target 'modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/cuda_compile_1_generated_nvidiaOpticalFlow.cu.o' failed
make[2]: *** [modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/cuda_compile_1_generated_nvidiaOpticalFlow.cu.o] Error 1
CMakeFiles/Makefile2:18502: recipe for target 'modules/cudaoptflow/CMakeFiles/opencv_cudaoptflow.dir/all' failed
make[1]: *** [modules/cudaoptflow/CMakeFiles/opencv_cudaoptflow.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....

I guess you didn't intend to break this but it seems that this code has no longer compatibility with CUDA 8.0.
Could you have a look on this issue please ? (I guess that's quicker to fix(

Best

[vchiluka5]

Hi @vchiluka5
Since this PR was merged, I see build failure on Jetson TX1 and TX2.
The build passes on Xavier, NX and Nano, which has CUDA 10.0 or later, but my TX1 and TX2 still has CUDA 8.0.

The error message is as following.

[ 57%] Building NVCC (Device) object modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/cuda_compile_1_generated_nvidiaOpticalFlow.cu.o
/opencv_contrib/modules/cudaoptflow/src/cuda/nvidiaOpticalFlow.cu(76): error: no instance of overloaded function "surf2Dwrite" matches the argument list
            argument types are: (short2, cudaSurfaceObject_t, unsigned long, int, cudaSurfaceBoundaryMode)

1 error detected in the compilation of "/tmp/tmpxft_00001b35_00000000-7_nvidiaOpticalFlow.cpp1.ii".
CMake Error at cuda_compile_1_generated_nvidiaOpticalFlow.cu.o.Release.cmake:279 (message):
  Error generating file
  /opencv/build/modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/./cuda_compile_1_generated_nvidiaOpticalFlow.cu.o


modules/cudaoptflow/CMakeFiles/opencv_cudaoptflow.dir/build.make:70: recipe for target 'modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/cuda_compile_1_generated_nvidiaOpticalFlow.cu.o' failed
make[2]: *** [modules/cudaoptflow/CMakeFiles/cuda_compile_1.dir/src/cuda/cuda_compile_1_generated_nvidiaOpticalFlow.cu.o] Error 1
CMakeFiles/Makefile2:18502: recipe for target 'modules/cudaoptflow/CMakeFiles/opencv_cudaoptflow.dir/all' failed
make[1]: *** [modules/cudaoptflow/CMakeFiles/opencv_cudaoptflow.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....

I guess you didn't intend to break this but it seems that this code has no longer compatibility with CUDA 8.0.
Could you have a look on this issue please ? (I guess that's quicker to fix(

Best

Optical flow API supports from CUDA 10 and above. You can add an exception which doesnt include nvidia_optical_flow using
-DHAVE_NVIDIA_OPTFLOW=0.