The shortest memo for building a python environment that takes less than 5 minutes to work.

Python is already on the mac, and I felt like I was doing it, but it's not. There are many situations where you need to switch between 2 and 3 in python. If you want to do python, you can create a conda environment. I don't use virtualenv or docker or anything like that. Only conda from pyenv is used here. conda is enough.

Put pyenv

ubuntu

git clone https://github.com/yyuu/pyenv.git ~/.pyenv

mac

brew install pyenv

win doesn't need pyenv. To conda.

Through pyenv

ubuntu

bashrc

#pyenv                                                                                                                                                        
export PYENV_ROOT=$HOME/.pyenv
export PATH=$PYENV_ROOT/bin:$PATH
eval "$(pyenv init -)"

mac

bashrc

#pyenv                                                                                                                                                        
eval "$(pyenv init -)"

Verification

pyenv versions

Install anaconda or miniconda

mac linux

pyenv install miniconda3-3.10.1
pyenv global miniconda3-3.10.1

win Download what you like https://www.continuum.io/downloads

Through conda

bashrc

export PATH="$PYENV_ROOT/versions/miniconda3-3.10.1/bin:$PATH"

Create a python environment with conda

conda create -n py35 python=3.5
conda create -n py37 python=3.7

Verification

conda info --e

Use an environment named py35

mac linux

source activate py35

win

activate py35

You can use the environment when (py35) is prompted. Sometimes (py35) appears, but I haven't put it in the environment. If you think the movement is strange, deactivate it once to exit the environment and enter again. If it is still strange, uninstall pyenv and re-enter the environment.

windows should try using conda once

I couldn't enter with pip, but I entered with conda.

conda install -c conda-forge keras

#3.1.0 series
conda install -c menpo opencv3=3.1.0 -n [Environment name]
#3.2.0 series
#3 at the time of writing.1.MSER for 0 series_There seems to be a bug in create, 3.2.I wrote that it was fixed at 0, so I wanted to put it back in. But I didn't enter.
#conda install -c conda-forge opencv=3.2.0
conda install -c conda-forge dlib=19.4
conda install scikit-image
conda install -c conda-forge ffmpeg

2018/6/15: 3.2.0 cannot be used Problem onset

The build is troublesome. The C ++ environment I created is likely to break. ..

And it doesn't disappear. opencv3 doesn't disappear from conda.

conda list | grep open
opencv3                   3.1.0                    py35_0    menpo

sudo conda install -c conda-forge opencv -n [environment]
sudo conda uninstall -c menpo opencv -n [environment]

sudo conda uninstall -c menpo opencv3
I don't know.

Switch the environment.

conda uninstall -c conda-forge opencv=3.2.0

>>> import cv2
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: No module named 'cv2'

I don't know anymore. Let's leave it for a while.

https://stackoverflow.com/questions/42310099/failed-to-run-conda-install-c-menpo-opencv3-3-2-0-in-windows7/42535587?noredirect=1

opencv

It's hard to build. Enter with pip or conda.

pip install -U opencv-python

Notes on jupyter

Be sure to use -n to name the environment when conda create. If you install without -n, it will be strange.

ImportError: No module named cv2 or ImportError: No module named tensorflow If something goes wrong, let's recreate the environment.

Shortest memo for kernel switching with jupyter

conda create -n py36 python=3.6
source activate py36
conda install -n py36 jupyter opencv
conda install notebook ipykernel
ipython kernel install --user --name py36 --display-name py36
jupyter noteobok

See here.

Solution when kernel switching on Jupyter does not work

Coexistence of python2 and python3 in jupyter

Look here and set https://qiita.com/utahkaA/items/ae14623ae81f85bea4f5

If you put both miniconda2 and miniconda3 in pyenv, it's OK. Even if I switched from miniconda3 to python2, it was not displayed in jupyter.

When a jupyter kernel error or an unreliable error occurs.

It's a good idea to generate a pass and put the encrypted one in the config file.

...
File "/usr/lib/python2.7/socket.py", line 224, in meth
return getattr(self._sock,name)(*args)
error: [Errno 99] Cannot assign requested address

The permissions error is not fixed by pip.

https://stackoverflow.com/questions/27870003/pip-install-please-check-the-permissions-and-owner-of-that-directory A phenomenon that is rarely seen. There are times when it is useless to delete everything under .pyenv and reinsert it.

Like putting owner authority in such an atmosphere
$ sudo chown -R USERNAME /Users/USERNAME/Library/Logs/pip
$ sudo chown -R USERNAME /Users/USERNAME/Library/Caches/pip

--Like putting a user
pip install --user <package name>

I want to run jupyter remotely from the server

Make settings here https://qiita.com/Miggy/items/5466a2c1e968602f3ebe

Output ~ / .jupyter / jupyter_notebook_config.py Set the password to sha1 with ipython, write the contents to jupyter_notebook_config.py or set a little.

After that, it should work if you open the port in each cloud.

For gcp https://towardsdatascience.com/running-jupyter-notebook-in-google-cloud-platform-in-15-min-61e16da34d52

Find out the Python import library path

Opencv Note where you are calling from.

import cv2
cv2.__file__
import imp
imp.find_module('cv2')

In my case, I went to see below .pyenv, which was as expected.

reference

https://github.com/jupyter/notebook/issues/2213 http://jupyter-notebook.readthedocs.io/en/latest/security.html https://qiita.com/SaitoTsutomu/items/aee41edf1a990cad5be6