Mac环境下已成功安装cv2但运行代码时触发ImportError的问题求助

Hey, let's fix this OpenCV import error you're hitting on your MacBook Air. Looking at the error message:

ImportError: dlopen(..., mach-o file, but is an incompatible architecture (have 'arm64', need 'x86_64'))

This means your Python interpreter is running in x86_64 architecture (likely via Rosetta), but the OpenCV package you installed is built for arm64 (the native M-series chip architecture). The two can't work together, so we need to align their architectures.

Step 1: Confirm the architectures #

First, let's verify what architectures your Python and OpenCV are using:

1. Check your Python's architecture:

   python3 -c "import platform; print(platform.machine())"
   

   • If output is x86_64: Your Python is running via Rosetta
   • If output is arm64: You're using the native M-series Python

2. Check OpenCV's architecture:

   file /Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/cv2/cv2.cpython-310-darwin.so
   

   Look for the architecture info in the output (it should say arm64 right now, which doesn't match your Python).

Step 2: Fix the mismatch #

Choose the solution that matches your Python's architecture:

Option 1: If Python is x86_64 (Rosetta)

• Uninstall the current arm64 OpenCV:

  pip3 uninstall opencv-python -y
  

• Open a new Terminal with Rosetta enabled:
  Right-click Terminal in Launchpad → Show Info → Check "Open using Rosetta"
• Reinstall OpenCV in this Rosetta Terminal:

  pip3 install opencv-python
  

  This will install the x86_64 version of OpenCV that matches your Python.

Option 2: If Python is arm64 (native)

• Uninstall the mismatched OpenCV:

  pip3 uninstall opencv-python -y
  

• Make sure you're using a native Terminal (no Rosetta enabled), then reinstall OpenCV:

  pip3 install opencv-python
  

  This will pull the arm64-compatible version of OpenCV.

Bonus: Using Conda (if you have it installed)

Conda automatically handles architecture matching, so this is a foolproof option:

1. Create a new environment:

   conda create -n cv_env python=3.10 -y
   

2. Activate the environment:

   conda activate cv_env
   

3. Install OpenCV via conda-forge:

   conda install -c conda-forge opencv -y
   

Step 3: Verify the fix #

Run your test code again:

import cv2
img=cv2.imread("galaxy.jpg",1)
print(img)

• If you see a numpy array printed (not None), the import worked!
• If you get None, double-check that galaxy.jpg is in the same folder as your script.py, or use the full absolute path to the image.

Also, make sure VS Code is using the correct Python interpreter (check the bottom-left corner of the window to select the one you installed OpenCV into).

内容的提问来源于stack exchange，提问作者Akshit