OpenCV 3.2.0 build with CUDA 8.0, TBB, extraModule

Setting the develop environment

OpenCV 3.2.0 build with CUDA 8.0, TBB, extraModule

Introduction

• 현재 openCV3.2.0(최신버전, 2017.05.xx) 및 extraModule를 이용하여 cmake에서 빌드를 하는 방법은 많이 포스팅 되어있으나, 빌드가 제대로 되지 않아 포스팅 함
• openCV 2.4.10, openCV 2.4.13, openCV 3.2.0 등 x86, x64 버전에에 대해 visual studio (2010, 2013, 2015, 2017) 제각각 돌려보았으나, 각 버전별로 차이가 있음. 그러나 중요한 것은 build tools의 호환성 문제도 있지만, 초보자들에 있어 path 문제 및 버전을 제대로 맞추지 못하여 생기는 문제로 생각됨
• 아래는 다 정리하지는 못하였으나, 몇가지 유의사항을 살펴본다면 별다른 문제가 없을 것으로 생각됨(?)

environments

• Visual studio 2015 Community edition
• openCV 3.2.0 / http://opencv.org/releases.html
• openCV extra Module / https://github.com/opencv/opencv_contrib
• cmake 3.8.1 / https://cmake.org/download/
• CUDA 6.519 (x64) / https://developer.nvidia.com/cuda-toolkit-65
• additional) cuda archive link / https://developer.nvidia.com/cuda-toolkit-archive
• CUDA 8.0 (x64)
• Intel MKL(math kernel library) / https://registrationcenter.intel.com/en/forms/?productid=2558&licensetype=2

주의사항

- opencv extra module 버전 및 opencv source 버전이 맞지 않으면 build 되지 않음(* opencv source 파일로 download 할 것)

• setting on cmake 3.8.1

  • check List
    • with_tbb
    • with_cublas
    • MKL_WITH_TBB
    • CUDA_FAST_MATH
  • uncheck list
    • build_docs
    • build_examples
    • build_package
    • build_perf_tests
    • build_tbb (automatic uncheck from system)
    • build_tests
    • build_with_static_crt

• setting the CUDA path

  • cuda_toolkit_root_dir : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v6.5
    or
  • cuda_toolkit_root_dir : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0

• setting the tbb path
  

  • TBB_ENV_INCLUDE :
    • D:/SDK/_opencv_build/opencv_extra_tbb/tbb2017_20170412oss_win/tbb2017_20170412oss/include
  • TBB_ENV_LIB :
    • D:/SDK/_opencv_build/opencv_extra_tbb/tbb2017_20170412oss_win/tbb2017_20170412oss/lib/intel64/vc14/tbb.lib
  • TBB_ENV_LIB_DEBUG :
    • D:/SDK/_opencv_build/opencv_extra_tbb/tbb2017_20170412oss_win/tbb2017_20170412oss/lib/intel64/vc14/tbb_debug.lib
  • TBB_VER_FILE :
    • D:/SDK/_opencv_build/opencv_extra_tbb/tbb2017_20170412oss_win/tbb2017_20170412oss/include/tbb/tbb_stddef.h
  • Note!
    • visual studio 2013 : vc12
    • visual studio 2015 : vc14
    • x86, x64를 구분하여 tbb dll library를 선택 할 것

• setting extra modules path
  

  • extra_modules :
    • D:\SDK_opencv_build\opencv_extra_modules\opencv_contrib-3.2.0_20170529\opencv_contrib-3.2.0\modules

• setting the MKL(math kernel library), IPP(integrated performance primitives)
  

• OS version에 맞춰서 download 할 것
• 이후, cmake 3.8.1 버전에서 자동으로 MKL library의 path를 설정함

• 자세한건, 아래의 path (C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows)로 들어가면 각각의 library path들을 확인 가능함

• Error of extra modules

OpenCVModule.cmake:260 CMake Error at cmake/OpenCVModule.cmake:260 (foreach): Syntax error in cmake code at E:/opencv/sources/cmake/OpenCVModule.cmake:260 when parsing string E:/opencv/sources/modules;EXTRA;E:\opencv\sources\modules\opencv_contrib-master\modules Invalid escape sequence \o Call Stack (most recent call first): modules/CMakeLists.txt:7 (ocv_glob_modules) Configuring incomplete, errors occurred! See also "E:/opencv/new_build/CMakeFiles/CMakeOutput.log". See also "E:/opencv/new_build/CMakeFiles/CMakeError.log".

solve

- change the special character from "\" to "/"

Reference

[1] 다크프로그래머 blog, http://darkpgmr.tistory.com/160

[2] cvlecture, http://cvlecture.marearts.com/2017/04/opencv-opencv-32-cuda-tbb.html

* 그외 많은 블로그 혹은 구글링으로 나오는 자료들을 많이 참조하였습니다. 포스팅 전 따로 저장해두지 못하여 Reference를 남기지 못합니다.