Install opencv with conda

Question

This question is different from "How do I install Python OpenCV through Conda?" because that question was asked more than 5 years ago, when all packages had different versions. I tried ALL answers to that question, and neither worked. See the text of question for details.

How to install opencv with conda now, in July 2019? On a freshly installed anaconda, I did conda update conda (succesfully) then tried the following:

(base) C:\Users\mlearning>python
Python 3.7.3 (default, Mar 27 2019, 17:13:21) [MSC v.1915 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'cv2'
>>> import cv
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'cv'

(base) C:\Users\mlearning>conda install -c menpo opencv3
Collecting package metadata (repodata.json): done
Solving environment: (goes into infinite loop, after 10 minutes I pressed ^C)

(base) C:\Users\mlearning>conda install opencv
Collecting package metadata (repodata.json): done
Solving environment: failed
Initial quick solve with frozen env failed.  Unfreezing env and trying again.
Solving environment: failed

UnsatisfiableError: The following specifications were found to be incompatible with each other:

Package zipp conflicts for:
importlib_metadata -> zipp[version='>=0.3.2']
path.py -> importlib_metadata[version='>=0.5'] -> zipp[version='>=0.3.2']
anaconda==2019.03=py37_0 -> importlib_metadata==0.8=py37_0 -> zipp[version='>=0.3.2']
zipp
Package importlib_metadata conflicts for:
anaconda==2019.03=py37_0 -> importlib_metadata==0.8=py37_0
path.py -> importlib_metadata[version='>=0.5']
Package hdf5 conflicts for:
anaconda==2019.03=py37_0 -> h5py==2.9.0=py37h5e291fa_0 -> hdf5[version='>=1.10.4,<1.10.5.0a0']
hdf5
opencv -> hdf5[version='>=1.10.2,<1.10.3.0a0,>=1.8.18,<1.8.19.0a0,>=1.8.20,<1.9.0a0']
h5py -> hdf5[version='>=1.10.1,<1.10.2.0a0,>=1.10.2,<1.10.3.0a0,>=1.10.4,<1.10.5.0a0,>=1.8.18,<1.9.0a0']
pytables -> hdf5[version='>=1.10.1,<1.10.2.0a0,>=1.8.18,<1.8.19.0a0,>=1.8.18,<1.9.0a0']
Package mkl-service conflicts for:
mkl-service
anaconda==2019.03=py37_0 -> mkl-service==1.1.2=py37hb782905_5

(base) C:\Users\mlearning>conda install -c conda-forge opencv
Collecting package metadata (repodata.json): done
Solving environment: failed
Initial quick solve with frozen env failed.  Unfreezing env and trying again.
Solving environment: failed

UnsatisfiableError: The following specifications were found to be incompatible with each other:

Package hdf5 conflicts for:
anaconda==2019.03=py37_0 -> hdf5==1.10.4=h7ebc959_0
h5py -> hdf5[version='1.10.1,1.8.17|1.8.17.*,1.8.18|1.8.18.*,>=1.10.2,<1.10.3.0a0,>=1.10.3,<1.10.4.0a0,>=1.8.20,<1.9.0a0']
pytables -> hdf5[version='1.8.18|1.8.18.*,>=1.10.4,<1.10.5.0a0,>=1.8.18,<1.8.19.0a0,>=1.8.18,<1.9.0a0']
hdf5
Package mkl-service conflicts for:
mkl-service
Package importlib_metadata conflicts for:
importlib_metadata
path.py -> importlib_metadata[version='>=0.5']

Solution 1

I have had countless problems with installing opencv with conda This is my approach, create an env if you don't already have one

conda create -n py36 python=3.6
conda activate py36

Install opencv with pip NOT conda

pip install opencv-python

If your still having an issue, uninstall opencv, update ffmpeg

conda install -c conda-forge ffmpeg 

then rerun pip

UPDATE 2020

install pip with you env activate

conda install pip

verify pip is in your env

whereis
pip: /path/anaconda3/envs/your_env/bin/pip

Install opencv with pip

~/anaconda3/envs/your_env/bin/pip3 install opencv-python

Solution 2

Create a complete new environment and let conda deal with compatibilities:

conda create -n cv -c conda-forge opencv matplotlib

This will create a new environment named "cv" with python, opencv and matplotlib.

Today (Out. 11th, 2019) it installed:

• Python 3.7.3
• OpenCV 4.1.1
• MatPlotLib 3.1.1
• Numpy 1.17.2
• ... and all other dependencies.

Solution 3

directly run

pip install opencv-python

on spyder instead

Solution 4

I faced similar problem (only hdf5 conflicted). The reason is using incompatible version python.

How about creating new python3.6 environment before installing openCV? You can create new environment like this.

$ conda create -n py36 python=3.6

In addition you should also type this command to activate.

$ activate py36

Solution 5

The easiest way is:

conda install pip
pip install opencv-python

Have fun!

View more solutions

Comments

[merv]

Why not just create the env from the start around the fact that OpenCV is what you want to install? I.e., conda create -n opencv_env opencv. If you want Python 3.6 you can specify that as well: conda create -n opencv_env opencv python=3.6

[ashiba]

@merv There is no paticular reason. If I had to say, I didn't know the conda command's notation.

[user31264]

Had you read my post? Among other things, I did exactly this, and it didn't work.

[AMC]

I have had countless problems with installing opencv with conda Can you elaborate? Using pip for this should really be a last resort.

[AMC]

How about creating new python3.6 environment before installing openCV? You should be using a separate environment for each project anyway.

[AMC]

Can you elaborate on the issue you were having?

[Peteris]

Does not work at June 2020; results in 'ImportError: DLL load failed: The specified module could not be found'.