LoginSignup
7
6

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 5 years have passed since last update.

@shinsumicco

複数バージョンのOpenCVを共存させたい on macOS

Last updated at Posted at 2017-08-01

動機

コンピュータビジョン関連の開発をしたり，ライブラリを使っていると，OpenCVの2系と3系を切り替えたいときがある．(特にPython)

そこで，

  • virtualenvでOpenCVのバージョンを使い分けたいけど，どこのcv2.soからシンボリックリンクを作ればいいのか.
  • pyenvしか使っていないときにOpenCVのバージョンを使い分けたいとき，PYTHONPATHには何を設定すればいいのか．
  • g++でOpenCVを使ったプログラムをビルドする際は，cmakeするときのパスをどこに設定すればいいのか．
  • 簡単にOpenCV2系とOpenCV3系を切り替える方法．

をメモしておく．

brew unlink & brew linkで管理していたら事故を起こして不具合が出たので，別の方法を考案．
Linuxでもパスを適切に設定すれば，この方法は使えるはず．

準備

HomebrewでOpenCVの2系と3系をインストールする

Terminal
# まずインストールする
# Python2系に切り替えておく
$ brew update
$ brew tap homebrew/science
$ brew install cmake pkg-config jpeg libpng libtiff eigen openexr numpy
$ brew install tbb
$ brew install ffmpeg

# OpenCV2系のインストール
# Python2系バインドが一緒に入るはず
$ brew install opencv --with-tbb --with-ffmpeg —-with-vtk

# OpenCV3系のインストール
# Python2系バインドを入れる場合
$ brew install opencv3 —-with-contrib --with-ffmpeg --with-tbb —with-vtk
# Python3系バインドを入れる場合　(Python3に切り替えてから実行すること)
$ brew install opencv3 —-with-contrib --with-ffmpeg --with-python3 --with-tbb —with-vtk --without-python

必要であればunlinkしておく．

Terminal
$ brew unlink opencv opencv3

パスの設定

unlinkをしてあっても，パスを通せばコンパイル・リンク時に参照できる．

cmakeするときのオプション

  • -DオプションでOpenCV_DIRを設定 (OpenCVConfig.cmakeがあるディレクトリ)
    • OpenCV2 : -DOpenCV_DIR=/usr/local/opt/opencv/share/OpenCV/
    • OpenCV3 : -DOpenCV_DIR=/usr/local/opt/opencv3/share/OpenCV/

OpenCV_DIRが設定されている状態でfind_package(OpenCV REQUIRED)をすると，OpenCV_INCLUDE_DIRSとOpenCV_LIBSが設定されるので，

CMakeLists.txt
cmake_minimum_required(VERSION 2.8)

project(Sample)

add_executable(Project main.cpp)

find_package(OpenCV REQUIRED)

if(OpenCV_FOUND)
  include_directories(${OpenCV_INCLUDE_DIRS})
  target_link_libraries(Project ${OpenCV_LIBS})
endif(OpenCV_FOUND)

のcmakeが通るようになる．

cv2.soのシンボリックリンク元

OpenCV3のPython2バインドをHomebrewでインストール，OpenCV3のPython3バインドは自前ビルドしていると仮定する．
OpenCVモジュールを使いたいPythonのインストールパス以下のディレクトリ，/path/to/python/lib/python2.7/site-packagesに，以下のcv2.soからシンボリックリンクを貼る．
(OpenCV, Pythonのバージョンは適宜変更する)

  • OpenCV2 for Python2
    • /usr/local/Cellar/opencv/2.4.13.2/lib/python2.7/site-packages/cv2.so
    • または /usr/local/opt/opencv/lib/python2.7/site-packages/cv2.so
  • OpenCV3 for Python2
    • /usr/local/Cellar/opencv3/3.2.0/lib/python2.7/site-packages/cv2.so
    • または /usr/local/opt/opencv3/lib/python2.7/site-packages/cv2.so
  • OpenCV3 for Python3
    • /path/to/own-built-opencv3/lib/python3.5/site-packages/cv2.so

環境変数でシステムの設定自体をまとめて変える

g++関連の環境変数

  • OpenCV_DIR
    • OpenCV2 : /usr/local/opt/opencv/share/OpenCV/
    • OpenCV3 : /usr/local/opt/opencv3/share/OpenCV/
  • CPLUS_INCLUDE_PATH
    • OpenCV2 : /usr/local/opt/opencv/include/:$CPLUS_INCLUDE_PATH
    • OpenCV3 : /usr/local/opt/opencv3/include/:$CPLUS_INCLUDE_PATH
  • LIBRARY_PATH
    • OpenCV2 : /usr/local/opt/opencv/lib/:$LIBRARY_PATH
    • OpenCV3 : /usr/local/opt/opencv3/lib/:$LIBRARY_PATH
  • PATH
    • OpenCV2 : /usr/local/opt/opencv/bin:$PATH
    • OpenCV3 : /usr/local/opt/opencv3/bin:$PATH

Python関連の環境変数

OpenCV3のPython2バインドをHomebrewでインストール，OpenCV3のPython3バインドは自前ビルドしていると仮定する．
site-packagesにはcv2.soが存在するはず．
(/usr/local/opt/opencvや/usr/local/opt/opencv3に同様のディレクトリがあればそちらでもOKかもしれない)

  • PYTHONPATH
    • OpenCV2 for Python2 : /usr/local/Cellar/opencv/2.4.13.2/lib/python2.7/site-packages/:$PYTHONPATH
    • OpenCV3 for Python2 : /usr/local/Cellar/opencv3/3.2.0/lib/python2.7/site-packages/:$PYTHONPATH
    • OpenCV3 for Python3 : /path/to/own-built-opencv3/lib/python3.5/site-packages:$PYTHONPATH
      • 自前でOpenCV3のPython3バインドをビルドし，ホームディレクトリ以下などの適当な場所に置いておく

OpenCV_DIRもOpenCVのバージョンに合わせて正しく設定しておいたほうがよい．

実例

システムの設定ごと変えるために~/.bash_profileに以下のようなスクリプトを書いた
(メッセージが邪魔ならechoの行を消す)

~/.bash_profile
# OpenCV version setting
OpenCV_VERSION="3.2.0"
export OpenCV_DIR=""
if [ "$OpenCV_VERSION" = "2.4.13.2" ]; then
  # OpenCV 2.4.13.2
  echo "OpenCV 2.4.13.2"
  export OpenCV_DIR="/usr/local/opt/opencv/share/OpenCV/"
  export CPLUS_INCLUDE_PATH="/usr/local/opt/opencv/include/:$CPLUS_INCLUDE_PATH"
  export LIBRARY_PATH="/usr/local/opt/opencv/lib/:$LIBRARY_PATH"
  export PATH="/usr/local/opt/opencv/bin:$PATH"
  # Python bindings (only Python 2)
  export PYTHONPATH="/usr/local/Cellar/opencv/2.4.13.2/lib/python2.7/site-packages/:$PYTHONPATH"
elif [ "$OpenCV_VERSION" = "3.2.0" ]; then
  # OpenCV 3.2.0
  echo "OpenCV 3.2.0"
  export OpenCV_DIR="/usr/local/opt/opencv3/share/OpenCV/"
  export CPLUS_INCLUDE_PATH="/usr/local/opt/opencv3/include/:$CPLUS_INCLUDE_PATH"
  export LIBRARY_PATH="/usr/local/opt/opencv3/lib/:$LIBRARY_PATH"
  export PATH="/usr/local/opt/opencv3/bin:$PATH"
  # Python binding (Python 2 or Python 3)
  PYTHON_VERSION="2"
  if [ "$PYTHON_VERSION" = "2" ]; then
    export PYTHONPATH="/usr/local/Cellar/opencv3/3.2.0/lib/python2.7/site-packages/:$PYTHONPATH"
  elif [ "$PYTHON_VERSION" = "3" ]; then
    # Own-built library
    export PYTHONPATH="/Users/username/usr/lib/python3.5/site-packages:$PYTHONPATH"
  else
    echo "Invalid Python version"
  fi
else
  echo "Invalid OpenCV version"
fi

OpenCV_VERSIONとPYTHON_VERSIONを切り替えて再読み込みするとOpenCVのバージョンが切り替えられる．

Terminal
$ source ~/.bash_profile
OpenCV 3.2.0
$

切り替えられるか確認

g++

以下のC++コードをコンパイルする．

main.cc
#include <iostream>
#include <opencv2/core.hpp>

int main()
{
  std::cout << "version: " << CV_VERSION << std::endl;
}
Terminal
# OpenCV_VERSION="3.2.0"のとき
$ g++ main.cc -o main
$ ./main
version: 3.2.0

# OpenCV_VERSION="2.4.13.2"のとき
$ g++ main.cc -o main
$ ./main
version: 2.4.13.2

Python

import cv2して，cv2.__version__とcv2.__file__を確認する．

Terminal
# OpenCV_VERSION="3.2.0"のとき
$ python
Python 2.7.13 |Anaconda custom (x86_64)| (default, Dec 20 2016, 23:05:08) 
>>> import cv2
>>> cv2.__version__
'3.2.0'
>>> cv2.__file__
'/usr/local/Cellar/opencv3/3.2.0/lib/python2.7/site-packages/cv2.so'

# OpenCV_VERSION="2.4.13.2"のとき
$ python
Python 2.7.13 |Anaconda custom (x86_64)| (default, Dec 20 2016, 23:05:08) 
>>> import cv2
>>> cv2.__version__
'2.4.13.2'
>>> cv2.__file__
'/usr/local/Cellar/opencv/2.4.13.2/lib/python2.7/site-packages/cv2.so'

おわりに

頑張れば with CUDA/with non-CUDA も切り替えられそう？

7
6
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
Sign upLogin
7
6

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?