module 'cv2.dnn' has no attribute 'DictValue'

[thepowerfuldeez]

Hi! I've come to an issue with the latest opencv (4.8.0.76)
I think you've changed functionality which is now not backward compatible

Full traceback:

    import cv2
  File "/usr/local/lib/python3.8/dist-packages/cv2/__init__.py", line 181, in <module>
    bootstrap()
  File "/usr/local/lib/python3.8/dist-packages/cv2/__init__.py", line 175, in bootstrap
    if __load_extra_py_code_for_module("cv2", submodule, DEBUG):
  File "/usr/local/lib/python3.8/dist-packages/cv2/__init__.py", line 28, in __load_extra_py_code_for_module
    py_module = importlib.import_module(module_name)
  File "/usr/lib/python3.8/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/usr/local/lib/python3.8/dist-packages/cv2/typing/__init__.py", line 169, in <module>
    LayerId = cv2.dnn.DictValue
AttributeError: module 'cv2.dnn' has no attribute 'DictValue'```

Python=3.7

[asmorkalov]

DictValue class type is defined in cv2/dnn/__init__.pyi:60 and works with interpreter:

>>> q = cv2.dnn.DictValue(0)
>>> print(q)
< cv2.dnn.DictValue 0x7fefca480fd0>

Tested with Ubuntu, python 3.8 and opencv-python 4.8.0.76 from pip.

[VadimLevin]

Can't reproduce

Python 3.11.4 (main, Jun 20 2023, 16:51:49) [Clang 14.0.0 (clang-1400.0.29.202)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2
>>> import cv2.typing
>>> cv2.typing.LayerId is cv2.dnn.DictValue
True
>>> import platform
>>> platform.platform()
'macOS-12.6.3-x86_64-i386-64bit'
>>> cv2.__version__
'4.8.0'

[caoandong]

Can confirm that this issue happens in the latest release by simply running import cv2.

The previous version opencv-python==4.8.0.74 works fine,.

Maybe it has something to do with the way import work?

[VadimLevin]

@caoandong

Can you confirm that you observe an issue in a clean virtual environment?

python3 -m venv venv
source venv/bin/activate
pip install opencv-python==4.8.0.76
python
>>> import platform
>>> platform.platform()
'macOS-12.6.3-x86_64-i386-64bit'
>>> import cv2
>>> cv2.__version__
'4.8.0'

[caoandong]

OK yeah installing from the clean virtual env solves the problem.

I think the issue has something to do with pip uninstall opencv-python not uninstalling the previously installed version.

After searching and removing all previous installed versions of cv2 and reinstalling opencv-python it works.

[thepowerfuldeez]

Thank you! That should solve the problem

[alrafiabdullah]

I faced this issue while using YOLOV8. For my case, the docker container had opencv-python & opencv-python-headless installed (both old versions). I uninstalled them using pip and then installed opencv-python==4.8.0.74.

[lyheiyu]

For mac
source activate 'your env'
You can use a lower opencv version
pip install opencv-python==4.5.0XXX

[K-Ashik]

opencv-python==4.5.64 has solved the problems for me in macbook m1 slicon

[stillgreyfox]

I was having this issue as well.

1. I uninstalled all my opencv versions (found by pip list)
2. Manually deleted the cv2 in the system python's dist-packages directory
3. Finally, re-installing opencv-python-headless
4. Worked fine with the latest version 4.8.1.78

How to Fix Stale cv2 / opencv-python Confilcts

# uninstall all versions of opencv
pip uninstall $(pip list --format=freeze | grep opencv)

# remove the stale cv2 dist-packages directory
# (pay attention to your location and python version)
rm -rf /usr/local/lib/python3.10/dist-packages/cv2/

# I'm installing on a Docker image that I don't intend to give a display,
# so I use the -headless package to avoid installing libgl stuff
# install latest opencv-python
pip install opencv-python-headless

# (or opencv-python if you have the gl stuff)

python -c "import cv2; print(cv2.__version__)"
4.8.1

Bonus Context

I'm setting this up on the nvcr.io/nvidia/pytorch:23.10-py3 image, which comes with opencv 4.7.0 installed, and I too was installing requirements for YOLOv8 and anomalib, which bumped me to the latest opencv-python, and then I found out that I wanted opencv-python-headless, because I'm doing a headless docker system, and this package doesn't depend on libgl.so.1.

If you need to use opencv-python, it should work with this fix, just make sure you have the dependencies that fix the missing libgl.so.1 problem:

Fix for libgl.so.1 if you need to use opencv-python

libgl.so.1 opencv-python fix

apt install ffmpeg libsm6 libxext6 -y

[soulteary]

The above processing process was made into a small tool, thanks @stillgreyfox

# download the autofix tool
pip install opencv-fixer==0.2.5
# execute and wait is done
python -c "from opencv_fixer import AutoFix; AutoFix()"

https://github.com/soulteary/opencv-fixer