2

So I'm trying to decode a QR code image using code from this S.O. answer. Here's the adapted code:

import cv2

# Name of the QR Code Image file
filename = r"C:\temp\2021-12-14_162414.png"
# read the QRCODE image
image = cv2.imread(filename)
# initialize the cv2 QRCode detector
detector = cv2.QRCodeDetector()
# detect and decode
data, vertices_array, binary_qrcode = detector.detectAndDecode(image)
# if there is a QR code
# print the data
if vertices_array is not None:
    print("QRCode data:")
    print(data)
else:
    print("There was some error")

(This is the whole program; I was still experimenting.)

The PNG file itself is really small, just 43 KB in size, with resolution of 290x290 (24 bpp) containing just the QR Code.

However, I keep getting the error:

Traceback (most recent call last):
  File "C:/Repos/tesqr/decod-cv2.py", line 10, in <module>
    data, vertices_array, binary_qrcode = detector.detectAndDecode(image)
cv2.error: OpenCV(4.5.4) D:\a\opencv-python\opencv-python\opencv\modules\core\src\alloc.cpp:73: error: (-4:Insufficient memory) Failed to allocate 54056250000 bytes in function 'cv::OutOfMemoryError'

Why is alloc.cpp asking for 54 GB of RAM ???

I'm new with OpenCV, so please help me troubleshoot what went wrong.

The library I'm using is:

$ pip3 freeze | grep opencv
opencv-contrib-python-headless==4.5.4.60

the input image:

input image

Christoph Rackwitz
  • 11,317
  • 4
  • 27
  • 36
pepoluan
  • 6,132
  • 4
  • 46
  • 76
  • This link is relevant: https://stackoverflow.com/questions/70347191/how-to-stop-opencv2-qrcode-detectanddecode-from-causing-memory-issues#comment124363804_70347191 – stateMachine Dec 16 '21 at 04:13
  • 1
    Also, please, share the input image. – stateMachine Dec 16 '21 at 04:35
  • @stateMachine Here you go: https://drive.google.com/file/d/1zgZ5VH_Cqt1xPo3WMmYQ3zvQLl1de-Hd/view?usp=sharing – pepoluan Dec 16 '21 at 06:58
  • yeah that's an extreme code, and the image is one pixel per block. resize with INTER_NEAREST. that might make it usable. -- please check for [related issues](https://github.com/opencv/opencv/issues?q=qrcodedetector) and add this if it hasn't been reported yet – Christoph Rackwitz Dec 16 '21 at 07:54
  • the code contains binary data. perhaps that has something to do with it. -- are you okay with that picture/code being used as a test case? if not, could you produce such a QR code that causes the same issue? – Christoph Rackwitz Dec 16 '21 at 11:42
  • 1
    @ChristophRackwitz well, the QR Code is supposed to contain a Zip file (raw, not baseXX encoded). It is perfectly decodable using node.js libraries `zbar.wasm` or `jsqr`. Unfortunately I cannot generate other QR Codes, but the information within the Zip file is dummy info, so it's okay to be public. – pepoluan Dec 17 '21 at 02:22
  • 1
    I've just created an issue: https://github.com/opencv/opencv/issues/21287 – Christoph Rackwitz Dec 18 '21 at 10:49

1 Answers1

2

Short Answer:

Try WeChatQRCode

Long Answer:

There are several open memory issues about decoding with QRCodeDetector. I hope that in future versions it will be fixed. Meanwhile you can try WeChatQRCode also from cv2.

WeChatQRCode includes two CNN-based models: A object detection model and a super resolution model. Object detection model is applied to detect QRCode with the bounding box. super resolution model is applied to zoom in QRCode when it is small.

Your code modified:

import cv2

# Name of the QR Code Image file
filename = "2021-12-14_162414.png"
# read the QRCODE image
image = cv2.imread(filename)
# initialize the cv2 QRCode detector
detector =cv2.wechat_qrcode_WeChatQRCode(detector_prototxt_path = "detect.prototxt", detector_caffe_model_path = "detect.caffemodel", super_resolution_prototxt_path = "sr.prototxt", super_resolution_caffe_model_path = "sr.caffemodel")
# detect and decode
data, vertices_array = detector.detectAndDecode(image)
# if there is a QR code
# print the data
if vertices_array is not None:
    print("QRCode data:")
    print(data)
else:
    print("There was some error")

Output:

QRCode data:
('PK\x03\x04\n',)

As you can see, it needs prototxt and caffemodel files. You can find them here.

frankscitech
  • 51
  • 3