Develop Reference

How to make the opencv-contrib module work?

I followed this guide to install opencv (version 3.4.4) and the contrib modules because I want to work with the SIFT algorithm.
https://www.pyimagesearch.com/2018/05/28/ubuntu-18-04-how-to-install-opencv/
When I noticed, it was not working as expected after the installation, I deleted the build folder and tried again, but this did not work as well.
I imported like this:
import cv2 as cv
I tried to use SIFT (in python 3.6) in the following ways:
sift = cv.SIFT_create()
sift = cv.xfeatures2d.SIFT_create()
sift = cv.SIFT()
sift = cv.xfeatures2d.SIFT()
As this usually results in errors like this: Attribute Error: module cv2 has no attribute 'SIFT_create' (same thing happens for the other 3 options), I figure that I am either using it in an incorrect way or the installation process itself did not work properly.
After this, I found the pip install opencv-contrib-python and used it. With no results at all.
I would really appreciate some hints on how I can make opencv with contrib modules work.

You may have to say cv2.xfeatures2d_SIFT or similar. the modules of OpenCV don't necessarily map to python submodules.
Since the patent on SIFT expired in 2019, OpenCV moved it back into features2d (main repo) from xfeatures2d (opencv_contrib repo). Please use the most recent 3.4.x release, or 4.x.

AttributeError: partially initialized module 'cv2' has no attribute 'CascadeClassifier' (most likely due to a circular import)

I've been going round in circles for ages trying to figure this out. Why am I getting this attribute error? I've tried using absolute references, and get the same issue. PyCharm is also highlighting CascadeClassifier, cvtColor and COLOR_BGR2GRAY saying it cannot find reference in cv2.py. I'm not sure if more information is relevant to solving this problem, so please ask if more is needed.
import cv2
face_cascade = cv2.CascadeClassifier('read_only/haarcascade_frontalface_default.xml')
eye_cascade = cv2.CascadeClassifier('read_only/haarcascade_eye.xml')
grayed_images = []
for x in np_images:
gray_img = cv2.cvtColor(x, cv2.COLOR_BGR2GRAY)
grayed_images.append(gray_img)
print(x)

If I am correct, you are using an environment different from where you run.
Step1: In command line/terminal where you see the opencv-python when you run pip list:
run python command. Copy your code, check if it works (you can simply import cv2 alternatively)
If it works, my idea should be correct. Otherwise, there is something bigger.
Step 2: (Assuming step1 works.) In Pycharm, Under Run > Edit Configurations, change python interpreter to whichever interpreter is you are that has opencv.
Step2 better alternative: On Pycharm, open the terminal, pip install opencv-python. After that you should have the opencv.

Downgrade opencv to the version 4.5.5.64.
You can install opencv using
pip install opencv-python==4.5.5.64.
This works for me

Alternative for PIL?

I want to use one of the newest versions of Python, (3.5+), and I need to import Image. I use:
from PIL import Image
except this returns an error. I don't have the error on my right now but the important part is there is a problem with PIL (I have already determined that from my Python Discord server)
Can I use pillow? I think I have that installed. I would try it but I want to know how (and if) it's applicable. Thanks

Can I use pillow?
Sure. pillow is a successor project of PIL, as development on the latter was last continued in 2011.
The first thing mentioned in the official pillow documentation is how to import the Image class.

Python extension scikit-image error

I am trying to use scikit-image to do some research. The system is Windows 7 64bit, and the python version is 2.7, 64bit.
The first program I run is from: http://scikit-image.org/
The code is
from skimage import data, io, filter
image = data.coins() # or any NumPy array!
edges = filter.sobel(image)
io.imshow(edges)
io.show()
However, the problem happens, and the error message is:
C:\Python27\lib\site-packages\skimage\io_plugins\null_plugin.py:14:
RuntimeWarning: No plugin has been loaded. Please refer to
skimage.io.plugins()
for a list of available plugins.
warnings.warn(RuntimeWarning(message))
I believe that both Python and scikit-image are correctly installed. So, may I know what is wrong with it?
Any suggestion is appreciated. Many thanks.

Just had precisely the same issue on amazon linux. The issue is that skimage requires PIL, and PIL is not installed. In the latest skimage they added the dependency, but the version that I got installed with pip didn't have it yet.
The solution is
pip install Pillow
EDIT: and after that you will probably immediately face another issue, with skimage loading image, but not being able to read it (in particular, shape being empty tuple). Here's the solution
Why does scipy.ndimage.io.imread return PngImageFile, not an array of values

Image library for Python 3

What is python-3 using instead of PIL for manipulating Images?

The "friendly PIL fork" Pillow works on Python 2 and 3. Check out the Github project for support matrix and so on.

Christoph Gohlke managed to build PIL (for Windows only) for python versions up to 3.3: http://www.lfd.uci.edu/~gohlke/pythonlibs/
I tried his version of PIL with Python 3.2, and image open/create/pixel manipulation/save all work.

Qt works very well with graphics. In my opinion it is more versatile than PIL.
You get all the features you want for graphics manipulation, but there's also vector graphics and even support for real printers. And all of that in one uniform API, QPainter.
To use Qt you need a Python binding for it: PySide or PyQt4.
They both support Python 3.
Here is a simple example that loads a JPG image, draws an antialiased circle of radius 10 at coordinates (20, 20) with the color of the pixel that was at those coordinates and saves the modified image as a PNG file:
from PySide.QtCore import *
from PySide.QtGui import *
app = QCoreApplication([])
img = QImage('input.jpg')
g = QPainter(img)
g.setRenderHint(QPainter.Antialiasing)
g.setBrush(QColor(img.pixel(20, 20)))
g.drawEllipse(QPoint(20, 20), 10, 10)
g.end()
img.save('output.png')
But please note that this solution is quite 'heavyweight', because Qt is a large framework for making GUI applications.

As of March 30, 2012, I have tried and failed to get the sloonz fork on GitHub to open images. I got it to compile ok, but it didn't actually work. I also tried building gohlke's library, and it compiled also but failed to open any images. Someone mentioned PythonMagick above, but it only compiles on Windows. See PythonMagick on the wxPython wiki.
PIL was last updated in 2009, and while it's website says they are working on a Python 3 port, it's been 3 years, and the mailing list has gone cold.
To solve my Python 3 image manipulation problem, I am using subprocess.call() to execute ImageMagick shell commands. This method works.
See the subprocess module documentation.

You can use my package mahotas on Python 3. It is numpy-based rather than PIL based.

You want the Pillow library, here is how to install it on Python 3:
pip3 install Pillow
If that does not work for you (it should), try normal pip:
pip install Pillow

Depending on what is needed, scikit-image may be the best choice, with manipulations going way beyond PIL and the current version of Pillow. Very well-maintained, at least as much as Pillow. Also, the underlying data structures are from Numpy and Scipy, which makes its code incredibly interoperable. Examples that pillow can't handle:
You can see its power in the gallery. This paper provides a great intro to it. Good luck!

If you are on Python3 you can also use the library PILasOPENCV which works in Python 2 and 3. Function api calls are the same as in PIL or pillow but internally it works with OpenCV and numpy to load, save and manipulate images. Have a look at https://github.com/bunkahle/PILasOPENCV or install it with pip install PILasOPENCV. Not all PIL functions have been simulated but the most common functions work.