[python] Install opencv for Python 3.3

Is OpenCV still not available for Python 3.3 and do I really have to downgrade to Python 2.7 to use it? I didn't find much about it on the internet, only some posts from 2012 that OpenCV wasn't yet ported to be used in Python 3.x. But now it's 2014 and after trying to install the latest OpenCV 2.4.x and copying the cv2.pyd file to C:\Program Files (x86)\Python333\Lib\site-packages this still yields the error in Python IDLE:

>>> import cv2
Traceback (most recent call last):
  File "<pyshell#0>", line 1, in <module>
    import cv2
ImportError: DLL load failed: %1 ist keine zulässige Win32-Anwendung.

I had a lot of trouble getting opencv 3.0 to work on OSX with python3 bindings and virtual environments. The other answers helped a lot, but it still took a bit. Hopefully this will help the next person. Save this to build_opencv.sh. Then download opencv, modify the variables in the below shell script, cross your fingers, and run it (. ./build_opencv.sh). For debugging, use the other posts, especially James Fletchers.

Don't forget to add the opencv lib dir to your PYTHONPATH.

Note - this also downloads opencv-contrib, where many of the functions have been moved. And they are also now referenced by a different namespace than the documentation - for instance SIFT is now under cv2.xfeatures2d.SIFT_create. Uggh.

#!/bin/bash
# Install opencv with python3 bindings: https://stackoverflow.com/questions/20953273/install-opencv-for-python-3-3/21212023#21212023

# First download opencv and put in OPENCV_DIR

#
# Edit this section
#
PYTHON_DIR=/Library/Frameworks/Python.framework/Versions/3.4
OPENCV_DIR=/usr/local/Cellar/opencv/3.0.0
NUM_THREADS=8
CONTRIB_TAG="3.0.0"  # This will also download opencv_contrib and checkout the appropriate tag https://github.com/Itseez/opencv_contrib


#
# Run it
#

set -e  # Exit if error

cd ${OPENCV_DIR}

if  [[ ! -d opencv_contrib ]]
then
    echo '**Get contrib modules'
    [[ -d opencv_contrib ]] || mkdir opencv_contrib
    git clone [email protected]:Itseez/opencv_contrib.git .
    git checkout ${CONTRIB_TAG}
else
    echo '**Contrib directory already exists. Not fetching.'
fi

cd ${OPENCV_DIR}

echo '**Going to do: cmake'
cmake -D CMAKE_BUILD_TYPE=RELEASE \
    -D CMAKE_INSTALL_PREFIX=/usr/local \
    -D PYTHON_EXECUTABLE=${PYTHON_DIR}/bin/python3 \
    -D PYTHON_LIBRARY=${PYTHON_DIR}/lib/libpython3.4m.dylib \
    -D PYTHON_INCLUDE_DIR=${PYTHON_DIR}/include/python3.4m \
    -D PYTHON_NUMPY_INCLUDE_DIRS=${PYTHON_DIR}/lib/python3.4/site-packages/numpy/core/include/numpy \
    -D PYTHON_PACKAGES_PATH=${PYTHON_DIR}lib/python3.4/site-packages \
    -D OPENCV_EXTRA_MODULES_PATH=opencv_contrib/modules \
    -D BUILD_opencv_legacy=OFF  \
    ${OPENCV_DIR}


echo '**Going to do: make'
make -j${NUM_THREADS}

echo '**Going to do: make install'
sudo make  install

echo '**Add the following to your .bashrc: export PYTHONPATH=${PYTHONPATH}:${OPENCV_DIR}/lib'
export PYTHONPATH=${PYTHONPATH}:${OPENCV_DIR}/lib

echo '**Testing if it worked'
python3 -c 'import cv2'
echo 'opencv properly installed with python3 bindings!'  # The script will exit if the above failed.

Whether or not you install opencv3 manually or from Gohlke's whl package, I found the need to create/edit the file cv.py in site_packages as follows to make compatable with old code:

import cv2 as cv

Spent 3 hours trying the various options on Ubuntu 14.04LTS mentioned here and in another referenced tutorial to no avail. For a while tried with OpenCV3.0.0 but eventually switched to 3.1.0. The following worked:

cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D PYTHON3_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.4m.so \
-D PYTHON3_EXECUTABLE=/usr/bin/python3.4m \
-D PYTHON3_INCLUDE_DIR=/usr/include/python3.4m/ \
-D PYTHON_INCLUDE_DIR=/usr/include/python3.4m/ \
-D PYTHON3_INCLUDE_DIRS=/usr/include/python3.4m/ \
-D PYTHON_INCLUDE_DIRS=/usr/include/python3.4m/ \
-D BUILD_opencv_python3=ON \
.

output:

--   OpenCV modules:
--     To be built:                 core flann imgproc ml photo video imgcodecs shape videoio highgui objdetect superres ts features2d calib3d stitching videostab python3
--     Disabled:                    java world
--     Disabled by dependency:      -
--     Unavailable:                 cudaarithm cudabgsegm cudacodec cudafeatures2d cudafilters cudaimgproc cudalegacy cudaobjdetect cudaoptflow cudastereo cudawarping cudev python2 viz

--   Python 3:
--     Interpreter:                 /usr/bin/python3.4m (ver 3.4.3)
--     Libraries:                   /usr/lib/x86_64-linux-gnu/libpython3.4m.so (ver 3.4.3)
--     numpy:                       /usr/lib/python3/dist-packages/numpy/core/include (ver 1.8.2)
--     packages path:               /usr/lib/python3/dist-packages
-- 
--   Python (for build):      

And with virtualenv used the following cmake options:

cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=$VIRTUAL_ENV \
-D PYTHON3_EXECUTABLE=$VIRTUAL_ENV/bin/python3 \
-D PYTHON3_PACKAGES_PATH=$VIRTUAL_ENV/lib/python3.4/site-packages \
-D PYTHON3_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.4m.so \
-D PYTHON3_INCLUDE_DIR=$VIRTUAL_ENV/include/python3.4m \
-D PYTHON_INCLUDE_DIR=$VIRTUAL_ENV/include/python3.4m \
-D PYTHON3_INCLUDE_DIRS=$VIRTUAL_ENV/include/python3.4m \
-D PYTHON_INCLUDE_DIRS=$VIRTUAL_ENV/include/python3.4m \
-D BUILD_opencv_python3=ON \
.

If you have issues with ffmpeg includes add the following to remove video support:

-D WITH_FFMPEG=OFF \
-D WITH_GSTREAMER=OFF \
-D WITH_V4L=OFF \
-D WITH_1394=OFF \

Also take heed of the warning from cmake about using make clean. If you ran make clean you might have to uncompress the original package anew. Cmake is dead, long live the Cmake

Use the pip application. On windows you find it in Python3/Scripts/pip.exe and On Ubuntu you can install with apt-get install python3-pip. and so, use the command line:

pip3 install --upgrade pip

pip3 install opencv-python

On Windows use only pip.exe instead pip3

I know this is an old thread, but just in case anyone is looking, here is how I got it working on El Capitan:

brew install opencv3 --with-python3

and wait a while for it to finish.

Then run the following as necessary:

brew unlink opencv

Then run the following as the final step:

brew ln opencv3 --force

Now you should be able to run import cv2 no problem in a python 3.x script.

Just in case you found that pip3 install opencv-python takes too long for you, you can set number of build threads:

export MAKEFLAGS="-j8"
pip3 install opencv-python --no-cache-dir

(--no-cache-dir will ignore previous build)

For Ubuntu - pip3 install opencv-python sudo apt-get install python3-opencv

You can use the following command on the command prompt (cmd) on Windows:

py -3.3 -m pip install opencv-python

I made a video on how to install OpenCV Python on Windows in 1 minute here:

https://www.youtube.com/watch?v=m2-8SHk-1SM

Hope it helps!

A full instruction relating to James Fletcher's answer can be found here

Particularly for Anaconda distribution I had to modify it this way:

 cmake -D CMAKE_BUILD_TYPE=RELEASE \
    -D CMAKE_INSTALL_PREFIX=/usr/local \
    -D PYTHON3_PACKAGES_PATH=/anaconda/lib/python3.4/site-packages/ \
    -D PYTHON3_LIBRARY=/anaconda/lib/libpython3.4m.dylib \
    -D PYTHON3_INCLUDE_DIR=/anaconda/include/python3.4m \
    -D INSTALL_C_EXAMPLES=ON \
    -D INSTALL_PYTHON_EXAMPLES=ON \
    -D BUILD_EXAMPLES=ON \
    -D BUILD_opencv_python3=ON \
    -D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib/modules ..

Last line can be ommited (see link above)

If you're down here... I'm sorry the other options didn't workout. Try this:

conda install -c menpo opencv3

from Step 1 of Scivision's Tutorial. If that doesn't work, then go on to Step 2:

(Windows only) OpenCV 3.2 pip install

Download OpenCV .whl file here. The packages that mention contrib in their name include OpenCV-extra packages. For example, assuming you have Python 3.6, you might download opencv_python-3.2.0+contrib-cp36-none-win_amd64.whl to get the OpenCV-extra packages.

Then, from Command Prompt:

pip install opencv_python-3...yourVersion...win_amd64.whl

Note that the ...win_amd64.whl wheels packages from step 2 in that tutorial are meant for AMD chips.

Someone has published a docker container / file for this:

https://github.com/vipul-sharma20/docker-opencv3-python3

https://hub.docker.com/r/vipul20/docker-opencv3-python3/~/dockerfile/

You can pull the image directly from docker hub or follow the instructions in the Dockerfile to install.

Here a solution for (I believe as seen by 'cp34' on the link below) Python 3.4.

Go to to http://www.lfd.uci.edu/~gohlke/pythonlibs/#opencv.

Download the appropriate .whl.

Go to the directory where the .whl is saved.

Use pip to install the .whl. e.g. pip install opencv_python-3.0.0-cp34-none-win_amd64.whl

Then just use import cv2.

I can't comment on midopa's excellent answer due to lack of reputation.

On a Mac I (finally) successfully installed opencv from source using the following commands:

cmake -D CMAKE_BUILD_TYPE=RELEASE 
-D CMAKE_INSTALL_PREFIX=/usr/local 
-D PYTHON_EXECUTABLE=/Library/Frameworks/Python.framework/Versions/3.4/bin/python3 
-D PYTHON_LIBRARY=/Library/Frameworks/Python.framework//Versions/3.4/lib/libpython3.4m.dylib 
-D PYTHON_INCLUDE_DIR=/Library/Frameworks/Python.framework/Versions/3.4/include/python3.4m 
-D PYTHON_NUMPY_INCLUDE_DIRS=/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages/numpy/core/include/numpy 
-D PYTHON_PACKAGES_PATH=/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages 
/relative/path/to/source/directory/

Then,

make -j8

change 8 for the number of threads your machine can handle, to speed things up

sudo make install

I added a PYTHONPATH environment variable to my ~/.bash_profile file so that Python could find cv2.so:

PYTHONPATH="${PYTHONPATH}:/usr/local/lib/python3.4/site-packages”
export PYTHONPATH

[For those using PyCharm, I had to go to Preferences > Project Structure > Add Content Root, and added the path to cv2.so's parent directory: /usr/local/lib/python3.4/site-packages]

This command got me past errors such as:

Could NOT find PythonLibs, by explicitly declaring the python library path

ld: can't link with a main executable for architecture x86_64
collect2: error: ld returned 1 exit status
make[2]: *** [lib/cv2.so] Error 1
make[1]: *** [modules/python2/CMakeFiles/opencv_python2.dir/all] Error 2
make: *** [all] Error 2

by explicitly pointing to the libpython3.4m.dylib

In terminal, check that it worked with:

$python3
>>> import cv2

It's all good if you don't get ImportError: No module named 'cv2'

This worked on a Macbook Pro Retina 15" 2013, Mavericks 10.9.4, Python 3.4.1 (previously installed from official download), opencv3 from source. Hope that this helps someone.

Yes support for Python 3 it is still not available in current version, but it will be available from version 3.0, (see this ticket). If you really want to have python 3 try using development version, you can download it from GitHub.

EDIT (18/07/2015): version 3.0 is now released and python 3 support is now officially available