[Zengyf-CVer]

@glenn-jocher
I found a solution to the problem mentioned in #7075, the reason is that my yolov5 project path includes Chinese, so cv2.imshow(str(p), im0) This line of code cannot display the window, resulting in the inability to display pictures and videos, I will fix this bug.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improved OpenCV functions for better Chinese character handling in file paths.

📊 Key Changes

• Added a wrapper function around cv2.imshow to prevent recursion errors.
• Extended functionality to encode file paths for proper handling of Chinese characters within the imshow function.
• Redefined cv2.imread, cv2.imwrite, and cv2.imshow to utilize the new, improved functions supporting Chinese character file paths.

🎯 Purpose & Impact

• 🚀 Purpose: The main goal is to enhance compatibility with Chinese characters in file paths when using OpenCV functions, mitigating issues faced by users with non-Latin alphabets in their file system.
• 🌍 Impact: This change makes the YOLOv5 more accessible and user-friendly for a global audience, particularly assisting Chinese-speaking developers and users in running image processing tasks without encountering errors related to file path encoding.

[glenn-jocher]

@Zengyf-CVer thanks for the PR! I think we should use the strategy I raised in #7139 (comment) for this

[glenn-jocher]

@Zengyf-CVer looks like unicode escape is faster than is_chinese, so makes sense just to always apply it rather than to check first:

p = 'acbadsfasdfasdfadslfkhadsfhjlasdfhljj.jpg'

%timeit p.encode('unicode_escape')
462 ns ± 3.47 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)
%timeit is_chinese(p)
1.15 µs ± 11.3 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

[glenn-jocher]

@Zengyf-CVer ok I've updated per my recommendation and replaced

-    cv2.imshow(p.encode('unicode_escape').decode(), im)
+    cv2.imshow(p.encode().decode('utf-8', 'ignore'), im)

Does this work for your use case? If not can you send me the image or string that is causing the problem?

[Zengyf-CVer]

@glenn-jocher
Still can't display properly, here is my video path and result:

# video path:
/home/zyf/文档/yolov5/video/202202241928.mp4
# run:
python detect.py --weights ./yolov5s.pt --source ./video/ --view-img

[glenn-jocher]

@Zengyf-CVer here are my --view-img results on MacOS. I renamed zidane.jpg as 文档.jpg. Not sure what to do.

path.encode('unicode_escape').decode()

path.encode('').decode('utf-8', 'ignore')

[Zengyf-CVer]

@glenn-jocher
My result is also like this, the code will convert Chinese to a specific encoding, but at least the picture can be displayed normally.

[Zengyf-CVer]

@glenn-jocher
This works fine on Ubuntu：

imshow_(path.encode('unicode_escape').decode(), im)

But this doesn't show up:

imshow_(path.encode().decode('utf-8', 'ignore'), im)

[glenn-jocher]

@Zengyf-CVer got it. I've reverted to imshow_(path.encode().decode('utf-8', 'ignore'), im). Should be good to merge once CI completes.

[glenn-jocher]

@Zengyf-CVer PR is merged. Thank you for your contributions to YOLOv5 🚀 and Vision AI ⭐