ImportError: cannot import name 'dnn_superres' for python example of super resolution with opencv

Question

I am trying to run an example for upscaling images from the following website: https://towardsdatascience.com/deep-learning-based-super-resolution-with-opencv-4fd736678066

This is the code I am using:

import cv2
from cv2 import dnn_superres

# Create an SR object
sr = dnn_superres.DnnSuperResImpl_create()

# Read image
image = cv2.imread('butterfly.png')

# Read the desired model
path = "EDSR_x3.pb"
sr.readModel(path)

# Set the desired model and scale to get correct pre- and post-processing
sr.setModel("edsr", 3)

# Upscale the image
result = sr.upsample(image)

# Save the image
cv2.imwrite("./upscaled.png", result)

I have downloaded the already trained model from the website, called "EDSR_x3.pb" and when I run the code I get the following error:

Traceback (most recent call last):
  File "upscale.py", line 2, in <module>
    from cv2 import dnn_superres
ImportError: cannot import name 'dnn_superres'

I now it seems like there is no such method or class, but I have already installed opencv and the contrib modules. Why do I get this error?

Answer 1

The key is in the documentation for opencv-python. dnn_superres is an extra module and requires you to install opencv-contrib-python

pip install opencv-contrib-python

Answer 2

Run the below commands into your terminal

pip install --upgrade opencv-contrib-python
pip install --upgrade opencv-python

Then, Restart your Jupyter notebok kernel or Reactivate your virtual environment so that the updated package can be used properly.

Answer 3

your opencv version should be opencv4.2.0+, the same question: https://github.com/opencv/opencv_contrib/issues/2544

solution:

pip install --upgrade opencv-python
pip install --upgrade opencv-contrib-python

Answer 4

I had the same problem with Python 3.6.9 and opencv 4.2.0, but after the upgrade to 4.3.0, the problem disappeared. If you have no problem upgrading the version, try 4.3.0.

Answer 5

As of this date, to add something to the currently accepted answer :

If you are doing it in your local system :

If you have

opencv-python

you need to uninstall it first, you can check if you have above package or not by the command :

pip show opencv-python

if you have it, uninstall it first by

pip uninstall opencv-python

and install

opencv-contrib-python

by

pip install opencv-contrib-python

by default it will download the latest version but still check the version by the show command above for opencv-contrib-python this time, just to check the version is above opencv4.2.0+, to force download a speicific version go through this : pip install package_name==version_number

If you are doing it in Google Colab :

opencv-python is already installed in colab but the version is 4.1.2, you can check it by :

import cv2
cv2.__version__

you don't need to uninstall it, you just need to install opencv-contrib-python but if you install by :

!pip install opencv-contrib-python

It won't install it, it would default to same preinstalled opencv

so run this :

!pip install opencv-contrib-python==

which will output all the available versions for colab, just select any version 4.3.x or above and run again, like:

!pip install opencv-contrib-python==4.3.0.36