Just found a solution for the "missing dll" issue

Solving the OpenCV `DLL load failed` Issue: A Step-by-Step Guide

If you’re using OpenCV with GPU acceleration (CUDA and cuDNN) and encountering the error ImportError: DLL load failed while importing cv2: The specified module could not be found, this article will walk you through diagnosing and fixing the issue based on a real-world example.

The Problem

While setting up OpenCV for Python 3.10 with GPU support, I encountered the following error when trying to import OpenCV (cv2):

Traceback (most recent call last):
  File "test.py", line 3, in <module>
    import cv2
  File "C:\Python310\lib\site-packages\cv2\__init__.py", line 181, in <module>
    bootstrap()
  File "C:\Python310\lib\site-packages\cv2\__init__.py", line 153, in bootstrap
    native_module = importlib.import_module("cv2")
  File "C:\Python310\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
ImportError: DLL load failed while importing cv2: The specified module could not be found.

Environment Details:

Python Version: 3.10.11
OpenCV Version: 4.7.0 (custom-built from source with CUDA support)
CUDA Version: 11.8
cuDNN Version: 8.9.7
Build Tool: Visual Studio 2019 (MSVC 14.29)
OS: Windows 10

Paths Configured:

PATH Environment Variable:

C:\DevelopmentTools\OpenCV\install\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin
C:\Program Files\NVIDIA\CUDNN\v8.9.7\bin

LIB Environment Variable:

C:\DevelopmentTools\OpenCV\install\x64\vc16\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\lib\x64
C:\Program Files\NVIDIA\CUDNN\v8.9.7\lib\x64

INCLUDE Environment Variable:

C:\DevelopmentTools\OpenCV\install\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\include
C:\Program Files\NVIDIA\CUDNN\v8.9.7\include

The Cause

This error is common when OpenCV cannot locate the shared libraries (DLLs) required during its runtime. Specifically:

Python 3.8+ DLL Loading Behavior:
- Starting with Python 3.8, the way Python loads DLLs has changed. It no longer automatically searches the system PATH environment variable for DLLs unless explicitly directed to do so.
- As a result, even though the required DLLs (like opencv_world470.dll, cudart64_118.dll, and cudnn64_8.dll) were present in the paths, Python didn’t load them.
OpenCV Loader Script:
- OpenCV’s loader script (cv2/__init__.py) dynamically handles paths for its binaries. However, it depends on either os.add_dll_directory() (Python 3.8+ feature) or a properly configured PATH.
- If these paths were not set correctly during runtime, OpenCV couldn’t load the required dependencies.

The Solution

Here’s the step-by-step process to resolve this issue:

1. Verify Environment Variables

Ensure the following paths are included in your environment variables:

Add OpenCV, CUDA, and cuDNN binary paths to the PATH:

C:\DevelopmentTools\OpenCV\install\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin
C:\Program Files\NVIDIA\CUDNN\v8.9.7\bin

Add library paths to the LIB environment variable:

C:\DevelopmentTools\OpenCV\install\x64\vc16\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\lib\x64
C:\Program Files\NVIDIA\CUDNN\v8.9.7\lib\x64

Add include paths to the INCLUDE environment variable:

C:\DevelopmentTools\OpenCV\install\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\include
C:\Program Files\NVIDIA\CUDNN\v8.9.7\include

2. Use `os.add_dll_directory()` for Python 3.8+

If you’re using Python 3.8 or newer, explicitly add these directories in your Python script before importing OpenCV:

import os

# Add OpenCV, CUDA, and cuDNN binary paths
os.add_dll_directory(r'C:\DevelopmentTools\OpenCV\install\bin')
os.add_dll_directory(r'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin')
os.add_dll_directory(r'C:\Program Files\NVIDIA\CUDNN\v8.9.7\bin')

import cv2
print(cv2.getBuildInformation())

This ensures Python explicitly knows where to find the required DLLs.

3. Recheck Required DLLs

Use a tool like Dependencies to analyze the cv2.cp310-win_amd64.pyd file. This tool shows if any required DLLs are missing. Common missing files include:

cudart64_118.dll (CUDA runtime)
cudnn64_8.dll (cuDNN library)
opencv_world470.dll (OpenCV core)

Ensure these files exist in the specified paths.

4. Update Visual C++ Redistributable

Install the latest Visual C++ Redistributable for Visual Studio 2019:

Download from Microsoft.

5. Test OpenCV

After completing the above steps, test OpenCV with a simple script:

import cv2
print(cv2.__version__)
print(cv2.getBuildInformation())

Conclusion

This issue stemmed from Python’s updated DLL loading behavior and OpenCV’s dependency on dynamically linked libraries. By ensuring that all required DLLs are in the correct paths and explicitly specifying these paths for Python, I was able to resolve the problem.

Key Takeaways

For Python 3.8+, use os.add_dll_directory() to manage DLL paths.
Ensure environment variables (PATH, LIB, INCLUDE) are configured correctly.
Use tools like Dependencies to diagnose missing DLLs.
Keep your CUDA, cuDNN, and Visual C++ Redistributable versions compatible with your OpenCV build.

Feel free to use this guide to resolve similar issues in your setup!

79308539